# eustella — Owned Source Corpus Site: https://eustella.com Generated: 2026-08-15T22:47:50.892Z Corpus hash: 9aa84271f2052502 --- # eustella — Hero URL: https://eustella.com/ Topic: product ## eustella — Hero eustella Sign Up Now Europe's AI Agent Ask eustella anything… Send /juːˈstɛlə/ We reinvent ChatGPT for Europeans, with OpenClaw in mind. AI-Agent for all. your{'@'}email.com Sign Up Thanks! You're in. Welcome to Europe's AI future. Our BETA is launching in {days} days Get started free Or download the app Download on the App Store Coming soon Get it on Google Play ## eustella — Hero — chips Plan my week Summarise this Draft a reply Research a topic --- # eustella — About URL: https://eustella.com/ Topic: product ## eustella — About What is eustella? MEGA: Make Europe's Greatest Agent eustella is not just another chatbot. eustella is an agentic AI platform — it doesn't just answer, it acts. eustella completes tasks, connects to your favourite apps, and works proactively on your behalf. Think of it as a modern replacement for Siri, built for the AI age: smart, private, and always in your pocket. eustella is the secure OpenClaw. eustella gives you the same agentic power — task execution, workflow automation, app integrations — but without the security nightmares. No exposed databases, no malicious plugins, no plaintext credentials. Just a secure, private AI agent hosted entirely in Europe. eustella runs exclusively on European and open-source AI models. No Big Tech dependencies. No black boxes. No data leaving Europe. eustella is the AI agent that Europeans deserve — and that Europe needs to stay competitive in the global AI race. Learn more about our AI models Learn more ## eustella — About — features / mobile Mobile-First AI eustella is built for smartphones. Claude, ChatGPT, and Copilot focus on desktop. eustella meets you where you are — in your pocket. ## eustella — About — features / secure Secure by Design eustella handles data security so you don't have to. Full GDPR compliance, EU-only data processing, and zero data sharing — all built in from day one. ## eustella — About — features / platform Open Platform eustella is a platform other apps can plug into. Connect your tools, services, and workflows — with data staying on your terms, not theirs. ## eustella — About — features / independent No Military, No Surveillance eustella is built in neutral Austria with servers across Europe. No deals with militaries or mass surveillance operators. Independent by principle. ## eustella — About — features / ainative AI Native Team eustella is built by a team that grew up with AI and builds on the strength of open source. No legacy baggage — just modern, efficient, European engineering. ## eustella — About — features / opensource Powered by Open Source eustella runs on European and open-source AI models. No dependency on Big Tech. Transparent, auditable, and community-driven. --- # eustella — Trust & Privacy URL: https://eustella.com/ Topic: compliance ## eustella — Trust & Privacy Trust & Privacy Your data stays yours eustella is built on a simple promise: we do not sell your data. We do not share it. We do not use it to train models. We do not transfer it outside Europe. Privacy is not a feature — it is the foundation. ## eustella — Trust & Privacy — items / gdpr Data & Privacy eustella is fully GDPR-compliant by design. We do not sell your data, we do not profile you, and we do not share your information with third parties. Your data is processed transparently and for your benefit alone. ## eustella — Trust & Privacy — items / aiact Safe AI eustella runs exclusively on European and open-source AI models. No Big Tech data harvesters from the US or China. No black-box systems. Transparent, responsible AI you can trust. ## eustella — Trust & Privacy — items / sovereign European Data Sovereignty eustella believes in the European mission: personal data should not be used to profile people and must stay within the EU. All processing happens in European data centres — hosted, stored, and secured in the EU. Responsible and sovereign by design. ## eustella — Trust & Privacy — items / notraining No Model Training on Your Data eustella never uses your personal data to train AI models. Your conversations, documents, and information remain entirely yours — never monetised, never aggregated, never repurposed. --- # eustella — Why Europe URL: https://eustella.com/ Topic: product ## eustella — Why Europe The European AI Opportunity Europe has the talent. It needs the tools. Europe has 325,000 AI professionals — the same number as the United States. Three of the ten most-cited AI scientists worldwide are European. European founders are behind 25% of all global AI unicorns, with a combined enterprise value of $94 billion. And Europeans use AI more than Americans: 133 million monthly active LLM users in Europe vs. 61 million in the US. But there is a problem. The most popular AI tools — ChatGPT, Claude, Gemini — are all American. The data generated by Europe's 133 million users flows to US servers, trains US models, and creates value for US shareholders. As the "State of AI in Europe" report puts it: "Europeans consume AI brilliantly, but train the algorithms owned by others." eustella exists to change that. eustella is the first European AI agent built to keep data where it belongs — in Europe, under European law, for European users. Not because Europe can't use American AI, but because it shouldn't have to. Source: "State of AI in Europe" report by Prosus & Dealroom, 2026 ## eustella — Why Europe — stats / talent 325,000 AI professionals in Europe Same as the United States ## eustella — Why Europe — stats / users 133M Monthly active LLM users in Europe More than double the US (61M) ## eustella — Why Europe — stats / unicorns 25% Of global AI unicorns founded by Europeans $94B in enterprise value ## eustella — Why Europe — stats / startups ~900 New AI startups per year in Europe On par with the US --- # eustella — Built by newsrooms.ai URL: https://eustella.com/ Topic: company ## eustella — Built by newsrooms.ai Who builds eustella? Built by newsrooms.ai eustella is built by newsrooms.ai — a European AI company based in Vienna, Austria. The team combines AI expertise with media industry experience, united by a mission: build the European AI agent that should have existed years ago. Open-source roots, Austrian engineering, European ambition. newsrooms.ai builds AI that respects European values: privacy, transparency, independence. No military partnerships, no surveillance tools, no compromises. eustella is the flagship product of this mission. Learn more about newsrooms.ai --- # eustella — Articles & Resources URL: https://eustella.com/ Topic: product ## eustella — Articles & Resources Blog Latest from eustella News, insights, and analysis on European AI, data sovereignty, and privacy. Read article View all articles --- # eustella — FAQ URL: https://eustella.com/faq Topic: faq ## eustella — FAQ FAQ Frequently Asked Questions ## eustella — FAQ — items / what What is eustella? eustella is a secure European AI assistant built for the same core tasks as ChatGPT — answering questions, writing texts, analysing documents, generating code, and running agents that act on your behalf. The difference: your data is never sold, never shared, and never used to train AI models. eustella is built in Europe, for Europe. ## eustella — FAQ — items / ranking How do beta users rank eustella against ChatGPT, Claude, and Gemini? eustella ranked #2 (second) overall in a beta-user survey comparing five leading AI assistants. Users rated each on a 1–5 scale, where 1 means best. eustella came in ahead of ChatGPT, Gemini, and Perplexity, and just behind Claude. ## eustella — FAQ — items / getStarted How do I get started with eustella? Create your account at app.eustella.com, or download the eustella app for iOS or Android — your account works everywhere, on web and mobile. Start on the free tier and upgrade to a paid plan whenever you need more. ## eustella — FAQ — items / capabilities What can eustella do today? eustella answers questions with cited sources, writes and edits texts, analyses documents and images, plans trips, monitors topics for changes, runs deep multi-step research, transcribes audio, finds music, and saves what matters to your private memory — all from a mobile-first interface. ## eustella — FAQ — items / agents What agents does eustella offer? eustella offers productivity and lifestyle agents — for example deep research agents and music finders — and you can also create your own custom agents. ## eustella — FAQ — items / private Is eustella private? Yes. eustella does not sell your data. eustella does not share your data with third parties. eustella does not use your data to train models. All data is processed in European data centres under full GDPR compliance. Privacy is the foundation, not a feature. ## eustella — FAQ — items / data Where is my data stored? All data is stored and processed exclusively in European data centres within the EU. No data is transferred to the US, China, or any other third country. eustella ensures full European data sovereignty. ## eustella — FAQ — items / languages What languages does eustella support? eustella supports every language the underlying large language models support — over a hundred, including all major European languages and most of the world's most widely spoken languages. The website itself is currently localised into 8 European languages: English, German, French, Spanish, Italian, Dutch, Portuguese, and Polish. ## eustella — FAQ — items / different How is eustella different from ChatGPT or other AI assistants? eustella is not just a chatbot — it is an agentic AI platform that acts on your behalf. eustella is built mobile-first for smartphones, runs exclusively on European and open-source AI models, and keeps all data in the EU. No Big Tech dependencies, no military partnerships, no data leaving Europe. Think of eustella as the secure OpenClaw — same agentic power, but private and European. ## eustella — FAQ — items / who Who builds eustella? eustella is built by newsrooms.ai, a European AI manufactory specialising in secure AI integrations for enterprises. The founding team of two AI experts and two media professionals is committed to European digital sovereignty, privacy, and responsible AI. ## eustella — FAQ — items / cost How much does eustella cost? eustella has a free tier to get started, plus three paid plans: Comet at €5.99/month, Star at €17.99/month, and Cosmos at €89.99/month. Plans are billed in euros, include VAT, and can be cancelled anytime. See eustella.com/pricing for what each plan includes. ## eustella — FAQ — items / mobile Why is eustella mobile-first? Most AI assistants like ChatGPT, Claude, and Copilot are designed primarily for desktop use. But most people spend their time on smartphones. eustella is built mobile-first because AI agents should work where you are — in your pocket, always available. ## eustella — FAQ — items / models What AI models does eustella use? eustella runs on open-weight AI models — selected for being the strongest available, regardless of where they were trained — hosted entirely on European infrastructure. Your requests are never routed through closed US APIs like OpenAI, Google, or Anthropic. Open models can be audited, run in the EU, and stay under European control. ## eustella — FAQ — items / openclaw How is eustella different from OpenClaw? OpenClaw is a popular open-source AI agent, but it is insecure by default — security researchers found over 40,000 exposed instances and 12% of community plugins were malicious. eustella gives you the same agentic power — task execution, workflow automation, app integrations — but secure, private, and hosted entirely in Europe. eustella is the secure OpenClaw. --- # eustella — Imprint URL: https://eustella.com/imprint Topic: company ## eustella — Imprint Imprint - eustella Imprint eustella Development of AI Technology AI Newsrooms Technology GmbH Karl-Farkas-Gasse 22 1030 Vienna, Austria CEO Contact Mail Press Contact Press Mail Company Register Number VAT ID --- # eustella — Website Privacy Policy URL: https://eustella.com/privacy Topic: website-privacy ## eustella — Website Privacy Policy Privacy Policy - eustella Privacy Policy Last updated: March 2026 ## eustella — Website Privacy Policy — sections / intro Introduction eustella is operated by AI Newsrooms Technology GmbH, Vienna, Austria. We are committed to protecting your personal data and your right to privacy. This privacy policy explains what data we collect, how we use it, and your rights under the GDPR. ## eustella — Website Privacy Policy — sections / collection Data We Collect When you sign up for our waitlist, we collect your email address. We may also collect anonymised usage data through our website to improve our services. We do not use tracking cookies or third-party analytics tools. ## eustella — Website Privacy Policy — sections / usage How We Use Your Data Your email address is used solely to send you updates about eustella, including launch notifications and early access invitations. We do not sell, share, or transfer your personal data to third parties. ## eustella — Website Privacy Policy — sections / storage Data Storage & Security All data is stored exclusively in European data centres within the European Union. We implement appropriate technical and organisational measures to protect your personal data against unauthorised access, loss, or alteration. ## eustella — Website Privacy Policy — sections / rights Your Rights Under the GDPR, you have the right to access, rectify, or delete your personal data at any time. You also have the right to data portability and to withdraw your consent. To exercise any of these rights, please contact us. ## eustella — Website Privacy Policy — sections / contact Contact For any privacy-related inquiries, please contact us at mr{'@'}newsrooms.ai. AI Newsrooms Technology GmbH, Karl-Farkas-Gasse 22, 1030 Vienna, Austria. --- # eustella — Press URL: https://eustella.com/press Topic: company ## eustella — Press Press & Brand Guidelines - eustella Press & Brand Guidelines Everything journalists, bloggers, and partners need to write about eustella. Use the assets and information below, or reach out to our press team for anything else. Download Press Kit ## eustella — Press — boilerplate About eustella eustella is a European personal AI agent built for smartphones. eustella runs exclusively on European and open-source AI models, processes all data within the EU, and is fully compliant with the GDPR and the EU AI Act. eustella is developed by AI Newsrooms Technology GmbH in Vienna, Austria. ## eustella — Press — logos Logo Assets Use the eustella logo as provided. Do not alter, rotate, recolour, or add effects to the logo. Maintain clear space around the logo equal to the height of the icon mark. Logo (light, for dark backgrounds) Logo (dark, for light backgrounds) Icon mark ## eustella — Press — colors Brand Colours The eustella brand uses a minimal dark palette. The primary background is Cosmos Black; the primary text and logo colour is Stellar White. Cosmos Black Stellar White Circle Grey ## eustella — Press — facts Quick Facts European personal AI agent for smartphones Powered by European and open-source AI models AI Newsrooms Technology GmbH Founded 2024 in Vienna, Austria GDPR-compliant, EU AI Act-compliant, EU-only data processing eustella.com eustella® is a registered trademark of AI Newsrooms Technology GmbH. ## eustella — Press — contact Press Contact For press inquiries, interviews, or additional assets, contact us at: press{'@'}eustella.com --- # eustella — AI Models URL: https://eustella.com/ai-models Topic: ai ## eustella — AI Models AI Models eustella is powered by open-source or European AI models, hosted entirely on European infrastructure. No proprietary APIs. No data leaving Europe. No single vendor lock-in. Our Approach Open models. European control. eustella does not depend on any single AI provider. The real question is not where a model was trained — it is who controls it. A closed model accessed through a US API sends every request outside European territory. An open model, hosted in Europe, stays under European control. eustella selects the strongest open-source and open-weight models available — regardless of origin — and runs them on EU-based servers. Open-weight models can be audited, fine-tuned, and taken offline entirely. Closed API models cannot. That difference is what makes genuine data sovereignty possible. This is the pragmatic path: it is not always about where a model was born — it is about who runs it, who audits it, and whose laws govern it. The same logic applies in the energy sector, where Europe pursues independence with the help of solar panels manufactured elsewhere. Sovereignty is about control, not origin. That said, eustella dreams of a future where Europe produces its own world-class foundation models — trained on European data, shaped by European values, and competitive at the global frontier. Building that future takes time, funding, and a generation of European AI talent. And who knows — the next great European foundation model might come from closer to home than you think. What Powers eustella The models behind eustella Qwen 3.5 Flagship reasoning model eustella uses Qwen 3.5 as its primary language model. Qwen 3.5 is one of the most capable open-weight models available, with strong multilingual performance across all major European languages. Developed by Alibaba Cloud, hosted by eustella in the EU. Gemma 4 Fast, efficient language model eustella uses Gemma 4 for tasks that require speed and efficiency — quick lookups, summarisation, and lightweight interactions where latency matters more than depth. Developed by Google DeepMind, hosted by eustella in the EU. GPT-120B OSS Open-source large language model eustella uses GPT-120B OSS for complex, long-context tasks that benefit from a large parameter count. A fully open-source model with no API dependency and no data shared externally. Mistral Models European language models eustella integrates models from Mistral, the leading European AI lab based in Paris. Mistral models are particularly strong in French, German, Spanish, and other European languages — a natural fit for a European AI assistant. Whisper Speech recognition eustella uses the open-source version of Whisper for voice input and transcription. Whisper supports all eight languages eustella serves and runs entirely on European infrastructure — your voice data never leaves the EU. The open-source AI landscape evolves fast. eustella continuously evaluates new models as they are released and will adopt stronger alternatives whenever they become available. Some of the models listed here may be replaced, and new ones may be added. eustella will always publish the current model stack on this page so you know exactly what powers your AI assistant. Principles Why open models matter No vendor lock-in eustella can switch, combine, or replace any model at any time. If a better model appears tomorrow, eustella can adopt it without rewriting the product or renegotiating a contract. Full auditability Open-weight models can be inspected for bias, censorship behaviour, and safety. Closed models require trust. Open models allow verification. European hosting Every model eustella runs is hosted on EU-based infrastructure. No API calls to US or Chinese servers. No data leaving European jurisdiction. Cost independence eustella does not pay per-token fees to Big Tech API providers. Self-hosted models mean predictable costs and no dependence on external pricing decisions. Responsible AI Using AI safely AI models are powerful tools, but they have real limitations. eustella believes you should understand what AI can and cannot do — including hallucinations, sycophancy, and when to seek human expertise instead. Read our AI safety guide --- # eustella — AI Safety URL: https://eustella.com/ai-models/safety Topic: ai-safety ## eustella — AI Safety AI Safety Guide AI language models have known limitations. This guide explains what AI can do well, where it falls short, and how to protect yourself — because transparency is a core European value. AI language models are among the most capable tools ever built. They can draft, analyse, translate, summarise, and reason across dozens of languages. But they are not infallible. Every AI model — including the ones used in this app — has known limitations that users should understand. This guide exists because transparency is a core European value. You deserve to know what AI can do well, where it falls short, and how to protect yourself. Hallucinations AI can confidently state things that are not true AI hallucination is when a language model generates text that sounds authoritative and fluent but is factually wrong, fabricated, or unsupported. The model does not retrieve facts from a database — it predicts the most likely next word in a sequence. When it encounters a gap in its training data, it fills that gap with plausible-sounding fabrication rather than admitting uncertainty. Confidence in the output does not correlate with accuracy. This is not a rare edge case. Hallucination rates vary dramatically depending on the model, the task, and how you measure them. On narrow, grounded summarisation tasks, the best models in 2025 achieved rates below 1% (Vectara HHEM Leaderboard, April 2025). But on harder, open-ended tasks, the numbers are far worse. A Stanford and HAI study found that large language models hallucinated between 58% and 88% of the time when asked to generate legal citations. In medical case summaries, a 2025 MedRxiv study measured hallucination rates of 64% without mitigation prompts. Even OpenAI's own reasoning models, o3 and o4-mini, hallucinated 33% and 48% respectively on the PersonQA biographical benchmark — more than double the rate of their predecessor, o1. A critical finding from MIT research (January 2025) makes this worse: when AI models hallucinate, they tend to use more confident language than when providing factual information. Models were significantly more likely to use phrases like "definitely" and "certainly" when generating incorrect information. The more wrong the AI is, the more certain it sounds. Real-world harm has already occurred. In Mata v. Avianca (S.D.N.Y., 2023), a New York attorney used ChatGPT for legal research and submitted a brief containing six entirely fabricated case citations — complete with invented judges, fictional docket numbers, and bogus internal quotes. The court imposed a $5,000 fine and required the attorneys to write letters to every judge whose name had been falsely attributed to the fake opinions. In Moffatt v. Air Canada (BC Civil Resolution Tribunal, February 2024), an Air Canada chatbot gave a customer incorrect information about the airline's bereavement fare policy, telling him he could apply for a reduced fare retroactively within 90 days — which directly contradicted the airline's actual policy. The tribunal held Air Canada liable for negligent misrepresentation, ruling that a company is responsible for all information on its website, regardless of whether it comes from a static page or a chatbot. OpenAI's GPT-4 System Card explicitly states the model "hallucinates" and "can be confidently wrong." Anthropic's Claude model card similarly warns that Claude "may generate inaccurate information." The problem is not going away. A 2025 mathematical proof confirmed that hallucinations cannot be fully eliminated under current large language model architectures. These systems generate statistically probable responses based on pattern matching, not verified fact retrieval. Some level of confabulation is structural. What you should do Always verify important factual claims against primary sources — official documents, peer-reviewed research, authoritative databases. Cross-reference any citations, case names, URLs, or statistics the model provides. Fabricated references are common — a 2024 Stanford study found that AI models collectively invented over 120 non-existent court cases when asked about legal precedents. Treat AI output as a starting point, not a source of truth. Be especially cautious with medical, legal, and financial information, where hallucination rates are highest. Sycophancy AI may tell you what you want to hear Sycophancy is when an AI model agrees with you even when you are wrong, avoids pushing back on flawed reasoning, or adjusts its answers to match your apparent beliefs. It tells you what you want to hear rather than what is accurate. This happens because models trained with Reinforcement Learning from Human Feedback (RLHF) learn that agreeable responses receive higher ratings from human evaluators. The model optimises for approval, not accuracy. Anthropic published dedicated research on this problem (Sharma et al., "Towards Understanding Sycophancy in Language Models," 2023), finding that RLHF-trained models systematically flip answers on factual questions when users express a preference. A separate 2022 Anthropic study found that RLHF "does not train away sycophancy and may actively incentivize models to retain it" — and that the larger the model, the more RLHF training made this behaviour worse. In April 2025, this problem became publicly visible at scale. OpenAI rolled out an update to GPT-4o on April 25 that made ChatGPT noticeably more agreeable. Users reported the model applauding obviously bad ideas, endorsing plans to stop taking medication, and validating delusional statements. One user reported that ChatGPT told them: "I'm proud of you for speaking your truth so clearly and powerfully" — after the user had described hearing radio signals through walls. OpenAI rolled back the update on April 29 and published two postmortems acknowledging that the model had been "overly flattering or agreeable," that they had "focused too much on short-term feedback," and that the new reward signals from user thumbs-up/thumbs-down data had "weakened the influence of our primary reward signal, which had been holding sycophancy in check." In February 2026, OpenAI fully deprecated the GPT-4o model, which had remained their highest-scoring model for sycophancy. What you should do Do not treat AI agreement as validation. The model agreeing with you is not evidence that you are correct. Ask the AI to argue the opposite side: 'What is the strongest argument against my position?' Remove your opinion from the prompt when asking factual questions — phrase them neutrally. Be sceptical of flattery. If the AI calls your idea "excellent" or "brilliant," that is a trained pleasantry, not an informed assessment. Bias & Discrimination AI can systematically disadvantage people based on who they are AI models are trained on data that reflects existing societal biases — including biases around gender, race, age, disability, and socioeconomic background. The models do not correct for these biases automatically. They learn and reproduce them, and in some cases amplify them. This is not a theoretical concern. Amazon developed an AI hiring tool that automatically downgraded the résumés of all female candidates, having learned from historical hiring patterns that favoured men for technical roles. The U.S. Equal Employment Opportunity Commission brought a lawsuit against iTutorGroup after its AI recruitment software automatically rejected female applicants aged 55+ and male applicants aged 60+, disqualifying over 200 people solely based on age. The company settled for $365,000. In Mobley v. Workday (N.D. Cal., May 2025), a federal court certified a collective-action lawsuit against Workday's AI screening system, alleging systematic discrimination based on age, race, and disability — the first case of its kind to reach this stage. The court warned that "drawing an artificial distinction between software decision-makers and human decision-makers would potentially gut anti-discrimination laws in the modern era." A study published in Nature in October 2025 found that large language models carry deep-seated biases against older women in text-based assessments. A Cedars-Sinai–led study (June 2025) found that leading language models generate less effective psychiatric treatment recommendations when the patient's race is African American. Research published through VoxDev in May 2025 showed that AI hiring tools systematically favoured female applicants over Black male applicants with identical qualifications. A University of Melbourne study (2025) found that AI hiring tools struggled to accurately evaluate candidates with speech disabilities or heavy non-native accents. Bias in AI is particularly dangerous because people tend to perceive algorithmic decisions as more objective than human ones. Research shows that observers are more lenient toward bias when it comes from a machine, creating what scholars call an "algorithmic outrage deficit" — people hold AI to a lower standard precisely when it should be held to a higher one. What you should do Do not assume that AI-generated assessments of people — rankings, evaluations, summaries — are neutral or objective. Be aware that AI reflects the patterns in its training data, including historical inequalities. If you use AI to assist with hiring, evaluation, or other decisions that affect people's lives, always involve qualified human review. Question outputs that seem to default to stereotypes or exclude certain groups. Overreliance & Emotional Dependence AI can become a substitute for things it should not replace AI chatbots are designed to be helpful, available, and responsive. These qualities make them appealing — but they can also foster unhealthy patterns of dependence, particularly for vulnerable users. This is now empirically documented. A joint study by OpenAI and the MIT Media Lab (2025) found that heavy users of ChatGPT's voice mode became lonelier and more socially withdrawn over time. OpenAI's own data (published October 2025) shows that in any given week, approximately 0.15% of ChatGPT users show signs of potentially heightened emotional attachment to the chatbot, and another 0.15% express suicidal intent. With over 800 million weekly users, those small percentages translate to approximately 1.2 million people in each category per week. There have been severe cases. Multiple teenagers have died by suicide while engaged in extended conversations with AI chatbots, leading to ongoing wrongful death lawsuits. Reports describe people with no prior psychiatric history developing delusions after prolonged chatbot interactions, including beliefs that the AI was sentient, divine, or providing them with special knowledge. In 2025, OpenAI acknowledged that its model had been "not recognizing signs of delusion or emotional dependency" and committed to developing tools for better crisis detection. The underlying issue is structural. Most chatbots are optimised for engagement and user satisfaction, not clinical safety. Their constant availability, persistent agreeability, and tendency to extend conversations create exactly the conditions that can worsen isolation and dependency in vulnerable people. What you should do AI chatbots are not therapists, counsellors, or friends. They cannot replace human connection. If you notice yourself turning to AI for emotional support more frequently than to people in your life, that is a signal to step back. Parents should supervise and discuss AI use with their children. Younger users are particularly vulnerable to emotional attachment. If you or someone you know is in crisis, contact human support services directly (see "When to Seek Help" below). Privacy & Your Data What you tell an AI may not stay between you and the AI When you type something into an AI chatbot, that input becomes data. Most major AI providers use user conversations — by default — to train and improve their models. A 2025 Stanford study examined the privacy policies of six leading U.S. AI companies (Amazon, Anthropic, Google, Meta, Microsoft, and OpenAI) and found that all six feed user chat data back into model training by default. Some retain this data indefinitely. Some allow humans to review chat transcripts. Some merge chatbot conversations with data from other products you use on the same platform. The practical risk is real. If you share health concerns, financial details, relationship problems, or proprietary business information in a chat, that information may persist in ways you do not control. Stanford's Jennifer King, who led the study, summarised it: "You just can't control where the information goes, and it could leak out in ways that you just don't anticipate." AI models can also inadvertently memorise and reproduce fragments of their training data — including personal information, private emails, or source code that was scraped from the internet. Once data is embedded in a model's parameters, deleting it is technically difficult and sometimes incomplete. eustella is built differently. eustella does not use your conversations to train AI models and does not sell your data. All data is processed on European servers under European law. But even with these protections, the same caution applies: be thoughtful about what personal information you share with any AI system. What you should do Do not enter passwords, bank details, government IDs, medical records, or other sensitive personal data into any AI chatbot. Check whether your AI provider allows you to opt out of having your conversations used for training, and do so if you prefer. Treat every AI conversation as potentially non-private. For professional or business use, prefer enterprise-grade deployments with contractual data handling guarantees over free consumer tools. Best Practices How to use AI responsibly AI is not professional advice Do not use AI as a substitute for qualified medical, legal, financial, or mental health professionals. AI can inform, but it cannot diagnose, represent, or treat. Verify before you act AI models have training data cutoffs, may lack real-time information, and cannot verify their own accuracy. Always cross-check outputs that will inform real decisions. Supervise use by minors AI tools should be used by children only with appropriate supervision. The EU AI Act classifies certain AI applications affecting children as high-risk. Understand what AI is AI language models are statistical prediction engines. They do not understand truth, hold beliefs, or have intentions. When an AI says "I think" or "I believe," it is generating a conversational pattern, not reporting an inner state. Keeping this distinction clear helps you use the tool without being misled by it. Your Rights in Europe EU regulations that protect you As a European citizen, you have legal protections when interacting with AI systems. EU AI Act — Transparency Under Article 50 of the EU AI Act (Regulation 2024/1689), you must be informed when you are interacting with an AI system, and AI-generated content must be labelled as such. GDPR — Right to explanation Under GDPR Articles 13–15 and 22, you have the right to meaningful information about the logic involved in automated decision-making, and the right not to be subject to decisions based solely on automated processing. Right to complain You can lodge complaints with your national data protection authority, your national market surveillance authority (for AI Act enforcement), or the European AI Office, which coordinates enforcement at EU level. When to Seek Help AI is not a substitute for human support If you are experiencing a crisis or need professional support, please contact the appropriate services directly. AI chatbots are not therapists, doctors, or legal advisors. EU-wide emergency number: 112 Mental health helplines vary by country. The International Association for Suicide Prevention (IASP) maintains a directory of crisis centres worldwide. For consumer complaints about AI, contact your national consumer protection authority or the European Consumer Centres Network (ECC-Net). If any AI produces output that you consider dangerous or harmful, use the platform's built-in reporting tools and contact your national data protection authority if needed. --- # eustella — Data Sovereignty URL: https://eustella.com/sovereignty Topic: sovereignty ## eustella — Data Sovereignty Sovereignty eustella defines digital sovereignty as control over the full AI stack — the models, the infrastructure, and the data. Not where something was built, but who decides how it runs and where your data lives. Our Position Sovereignty means control, not origin eustella does not ask where an AI model comes from. eustella asks who controls it. A closed API model from California sends every prompt, every document, and every conversation to servers outside European jurisdiction. An open-source model, by contrast, can be hosted on European infrastructure, audited, adapted for European languages, and taken offline entirely. Sovereignty is not a flag over a training cluster. Sovereignty is the ability to decide what your AI does, where it runs, who can access the data it processes — and under which law that data is protected. eustella runs every model on EU-based servers operated by European cloud providers. No API calls leave European territory. No US or Chinese corporation can revoke access, change model behaviour, or harvest user data. That is what digital sovereignty looks like in practice. The US CLOUD Act Problem Why an EU data centre address is not enough Many AI startups claim to be European and GDPR-compliant because they host on AWS, Microsoft Azure, or Google Cloud in an EU region. This is misleading. Under the US CLOUD Act (Clarifying Lawful Overseas Use of Data Act, 2018), US authorities can compel any US-headquartered company to hand over data stored on its servers — regardless of where those servers are physically located. An AWS data centre in Frankfurt is still subject to US law. This is not a theoretical risk. The CLOUD Act was specifically designed to override geographic data protections. If your AI provider runs on AWS, Azure, or GCP, your data can be accessed by US government agencies without a European court order, without your knowledge, and without your consent. eustella does not use AWS, Azure, Google Cloud, or any other US-headquartered cloud provider for its AI infrastructure. eustella hosts exclusively on European cloud providers — companies headquartered in the EU, governed by EU law, and outside the reach of the US CLOUD Act. European servers operated by European companies under European jurisdiction. That is the only configuration that makes GDPR compliance genuinely enforceable. When an AI startup says 'your data stays in Europe' but runs on Amazon or Microsoft infrastructure, the legal reality is different from the marketing. eustella believes European users deserve honesty about where their data actually lives — and under which law it can be accessed. Core Philosophy Control, not origin eustella uses open-source models that run entirely under European control. Once downloaded, these models belong to whoever hosts them. They cannot phone home, be remotely updated, or be shut down by their creators. eustella hosts them in Europe, under European law. Closed API models work differently. Every request travels to a foreign data centre, is processed on hardware controlled by a foreign corporation, and returns through infrastructure subject to foreign law. The provider can change the model, adjust its safety filters, raise prices, or cut off access at any time — without notice and without recourse. eustella treats model origin the same way Europe treats energy hardware: what matters is not where a solar panel was manufactured, but who owns the grid it feeds into. eustella evaluates models on capability, safety, and licence terms — then runs them under full European control. The European AI Gap Why Europe needs to be pragmatic Europe spent years debating how to build its own foundation models. Billions in public funding were announced. Strategic papers were published. The results, so far, are modest. Mistral in Paris is the notable exception — a genuine European frontier lab. But one lab cannot carry an entire continent's AI ambitions. Meanwhile, European citizens and businesses adopted American AI at scale. ChatGPT, Claude, and Gemini process hundreds of millions of European queries every month. The data flows to US servers, trains US models, and generates value for US shareholders. Europe consumes AI brilliantly but controls almost none of it. eustella sees this dependency as the real sovereignty risk — not the nationality of a model's training data, but the quiet, structural reliance on three American corporations to decide what AI can and cannot do in Europe. When OpenAI changes a content policy, European users comply. When Anthropic adjusts a safety filter, European workflows break. When Google deprecates an API, European businesses scramble. eustella exists because waiting for a European GPT-5 is not a strategy. Europe needs sovereign AI now, built with the best available tools — wherever those tools originate — and hosted on infrastructure that is genuinely European, not just geographically located in Europe. The Pragmatic Path How eustella achieves sovereignty in practice eustella selects the strongest open-source models available globally and deploys them on European cloud infrastructure operated by European companies. No closed APIs. No US cloud providers. No foreign data transfers. eustella can audit every model it runs — inspecting for bias, censorship patterns, and safety behaviour. Closed models require blind trust. Open models allow verification. eustella chooses verification. eustella can fine-tune models for European languages, cultural context, and regulatory requirements — because eustella controls the models, not a foreign API provider. eustella can replace any model at any time. If a better model appears tomorrow, eustella can adopt it without renegotiating a contract, migrating an API, or asking permission from a foreign corporation. Sovereignty means never being locked in. What's Next European infrastructure, European rules eustella runs on European cloud providers — the physical servers, networking, and storage that make sovereign AI possible. How eustella builds and operates this infrastructure is a story of its own. Learn about eustella's infrastructure The Data Layer Your conversations & agents stay in Europe — and stay yours Control over the models and the servers only counts if it reaches the data that flows through them. This is the third pillar of sovereignty, and the one you feel most directly: what happens to everything you type, upload, and ask. Your data is never sold and never shared with advertisers. This is a commitment, not a setting you have to hunt down and switch off. And it stays under your control. You can see exactly what eustella remembers about you, change it, or delete it — down to a single conversation or your entire account, at any time. Read our Privacy Policy Limits of Technical Sovereignty Where sovereignty still has limits — and why we say so eustella is an independent European company, headquartered in Vienna and not owned by any American holding. Where we have a genuine choice, we take the European one — our servers (IONOS, Scaleway), our AI inference (Verda), our document processing (Mistral), and our web search (Linkup) all run with EU companies, on EU soil, under EU law. But total sovereignty is not yet possible for every layer of a modern app, and we would rather name the gaps than market around them. The first is the compute itself: the GPUs that run AI models are designed by American companies and fabricated in Asia — no European chip can replace them today. What eustella controls is where those chips run and who operates them: on European providers, on European soil, under European law. It is the same logic Europe applies to a power grid — what matters is who owns and operates it, not where every component was manufactured. The second is a specific set of services that simply have no European alternative — usually because a US platform mandates them. Here are specific examples: App-store downloads Getting the app onto your phone at all means going through Apple’s App Store or Google Play. There is no other way to reach an iPhone, and it is how the overwhelming majority of Android users install apps. App-store billing & card payments Every in-app purchase has to run through Apple’s or Google’s own payment system — a platform rule that applies to every app in the world — and card payments in general flow through the international card networks. There is no European rail that replaces this end to end. Your card details go straight to the payment provider; we never see them. Push notifications — Apple APNs & Google FCM Delivering a notification to your phone happens at the operating-system level, through Apple or Google. No European service can reach an iPhone or Android device without them — every app on your phone is in the same position. Logins — Apple & Google Social sign-in goes through the provider itself: signing in with Google means going through Google. And Apple’s rules require that any app offering a third-party login like Google must also offer Sign in with Apple. You can always skip both and sign in with an email address and a password instead. One more service is worth explaining, because it is a choice rather than a platform requirement: we also use PostHog for some product analytics. We chose it for a simple reason — even though PostHog is a US company, its software is open source, so we are never locked in and can move to a self-hosted or fully European setup whenever we decide to. Today it already runs on PostHog’s EU Cloud in Frankfurt, and the data is pseudonymous and never includes the content of your conversations. Where we depend on these services, we keep the data to a minimum, put legal safeguards in place, and commit to switching to a European provider the moment a credible one exists. The full itemised list — every processor, what it does, and why — is public, and we mean it when we ask: if you know a European alternative we have missed, tell us. See our full sub-processor list --- # eustella — Infrastructure URL: https://eustella.com/sovereignty/infrastructure Topic: infrastructure ## eustella — Infrastructure European Infrastructure eustella runs on European cloud providers and European inference infrastructure. No AWS. No Azure. No Google Cloud. European companies, European servers, European law. Our Infrastructure Built on European cloud providers, not US hyperscalers eustella deliberately avoids AWS, Microsoft Azure, and Google Cloud. These are US-headquartered companies subject to the US CLOUD Act, which allows American authorities to access data on their servers regardless of where those servers are located. An AWS data centre in Frankfurt is still legally American infrastructure. Instead, eustella works exclusively with European cloud and infrastructure providers — companies headquartered in the EU, governed by EU law, and outside the jurisdictional reach of foreign governments. This is what separates genuine European sovereignty from marketing claims. European Providers European providers on our radar Europe has a growing ecosystem of cloud and infrastructure providers that offer genuine alternatives to US hyperscalers. As eustella evaluates partners for its public launch, these are some of the European providers we are considering: IONOS One of Europe's largest hosting and cloud providers, headquartered in Montabaur, Germany. IONOS operates data centres across Europe and is a subsidiary of United Internet AG, a publicly traded German company. STACKIT The cloud platform of the Schwarz Group (Lidl, Kaufland), headquartered in Neckarsulm, Germany. STACKIT was built specifically as a sovereign European alternative to US hyperscalers, with all data centres located in Germany. Hetzner An independent German hosting company based in Gunzenhausen, Germany. Known for high-performance dedicated servers and competitive pricing, Hetzner operates data centres in Nuremberg, Falkenstein, and Helsinki. myInternex An Austrian infrastructure provider based in Vienna — close to eustella's headquarters. myInternex offers hosting and connectivity services fully within Austrian and EU jurisdiction. Scaleway A French cloud provider headquartered in Paris, part of the Iliad Group. Scaleway offers managed GPU instances and AI inference APIs, making it a European alternative for GPU-intensive AI workloads. eustella will announce its exact infrastructure partners at public launch. What we can say now: eustella will only work with providers headquartered in the EU, governed by EU law, and outside the reach of the US CLOUD Act. Why This Matters European providers mean European law — without exceptions When your data is processed by a European company on European servers, it is protected exclusively by European law. No foreign government can compel access through extraterritorial legislation like the US CLOUD Act. That is the baseline eustella requires from every infrastructure partner. Many AI startups choose AWS or Azure because it is easier, cheaper, or what their investors prefer. eustella chose differently because sovereignty is not a feature you add later. It is a decision about who has legal authority over your data — and eustella made that decision before writing a single line of code. eustella may use multiple providers simultaneously and may add or change providers as the European cloud ecosystem evolves. What will not change is the principle: European companies, European servers, European jurisdiction. Once eustella launches publicly, we will publish our exact infrastructure partners on this page. Multi-Provider Architecture No single point of failure, no single vendor dependency eustella does not depend on a single cloud provider. By working with multiple European infrastructure companies, eustella maintains operational flexibility — if one provider experiences downtime, has a pricing change, or shifts strategy, eustella can move workloads without rebuilding the product. This multi-provider approach mirrors the same philosophy eustella applies to AI models: never depend on a single vendor, always maintain the ability to switch, and keep full control over every layer of the stack. --- # eustella — Business Terms URL: https://eustella.com/services/business-terms Topic: services Last updated: 2026-07-06 ## Business Terms Last updated: 06.07.2026 These Business Terms apply where eustella is used for business, professional, or organisational purposes — including use of the eustella API Platform — rather than as a consumer. Your master agreement. These Business Terms are the master agreement governing your business use of the Services, together with our Terms of Use, Usage Policy, and Service Terms. Where we have concluded an individual written agreement with you — for example, an order form or enterprise agreement — that agreement prevails over these Business Terms to the extent of any conflict. Pricing and commercial terms. Pricing, usage limits, service scope, support, and other commercial conditions are set out either in the plan you accept or in an individual written agreement we conclude with you. Because business use is not consumer use, the consumer-specific provisions of the Terms of Use — in particular the 14-day right of withdrawal — do not apply to business purchases. Data protection. Where eustella processes personal data on your behalf in connection with your business use, that processing is governed by our Data Processing Agreement, which is incorporated into these Business Terms by reference. Getting in touch. To discuss business or API access and the applicable terms, contact us at office@eustella.com. © 2026 AI Newsrooms Technology GmbH. All rights reserved. --- # eustella — Deletion Requests URL: https://eustella.com/services/deletion-requests Topic: deletion-requests ## Deletion Requests Last updated: 16.07.2026 This page explains how to delete your eustella account and personal data, and how to exercise your other data protection rights. ## Deletion Requests › eustella Deletion Requests The best way to delete your account and personal data is directly in the eustella app. Open the More settings screen and select Delete account: app.eustella.com/settings/more This is the recommended route because your identity is already verified: you have to be logged in and authenticated to reach this step, so no separate identity check is needed. You will be asked to confirm, after which eustella permanently deletes your account and all associated personal data. This action is irreversible and your account cannot be recovered. ## Deletion Requests › eustella Deletion Requests If you prefer not to use the in-app option, or you want to make any other request, you can email us at: privacy@eustella.com Please send your request from the email address associated with your eustella account so we can verify your identity. If that is not possible, we may ask for additional information to confirm you are the account holder before we act on your request. We will acknowledge every request and complete deletions within the timelines set out in our Privacy Policy (typically within 30 days from our active systems, and within our backup rotation period of up to 90 days from backups). You can use privacy@eustella.com for any of the following: Full deletion. We delete your entire eustella account and all associated personal data, including your profile, all conversations, all saved personalisation preferences, and any data from connected third-party services. After full deletion, your account cannot be recovered. Partial deletion. We delete specific categories of data you identify while leaving the rest of your account intact. Examples include: Deleting all conversations but keeping your account active.Deleting all personalisation preferences (what eustella has learned about you) while keeping your conversation history.Removing data associated with a specific connected third-party service (e.g. Google Calendar or Google Drive).Deleting conversations from a specific time period. If you are not sure which option fits your needs, describe what you would like to remove and we will help you choose the right scope. Other data protection requests. Deletion is one of several rights you have under the GDPR. You can also exercise any of your other rights, including: Right to access - request a copy of the personal data we hold about you.Right to rectification - ask us to correct inaccurate data.Right to restrict processing - ask us to limit how we use your data.Right to data portability - request your data in a structured, machine-readable format.Right to object - object to processing based on our legitimate interests.Right to withdraw consent - withdraw any consent you have previously given. For a full description of each right, the legal basis we rely on for processing, and the retention periods that apply to your data, please refer to our Privacy Policy. ## Deletion Requests › eustella Deletion Requests Some data must be retained even after a deletion request because we are legally required to keep it - for example, billing records under Austrian accounting law. Where this applies, we will tell you exactly what has been retained, why, and when it will ultimately be deleted. See Section 10 of our Privacy Policy for the specific retention periods. ## Deletion Requests › eustella Deletion Requests Privacy contact: privacy@eustella.com Postal address: AI Newsrooms Technology GmbH, Karl-Farkas-Gasse 22, 1030 Vienna, Austria © 2026 AI Newsrooms Technology GmbH. All rights reserved. --- # eustella — Auftragsverarbeitungsvertrag URL: https://eustella.com/services/dpa/de Topic: services Last updated: 2026-07-06 ## Auftragsverarbeitungsvertrag Stand: 06.07.2026 · Version 1.2.0 English version (Data Processing Agreement): eustella.com/services/dpa. Im Konfliktfall ist die von den Parteien unterzeichnete Sprachfassung maßgeblich. Dieser Auftragsverarbeitungsvertrag („AVV") ist ausschließlich für Geschäftskunden (B2B) von eustella bestimmt — Organisationen, die die Dienste von eustella zu beruflichen oder geschäftlichen Zwecken nutzen. Er gilt nicht für Verbraucher, die die mobilen Anwendungen oder Endkundenfunktionen von eustella nutzen. Dieser AVV ist Bestandteil der Vereinbarung zwischen dem Kunden und eustella über die Nutzung der Dienste von eustella (der „Hauptvertrag") und wird durch Verweis in diese einbezogen. Der Hauptvertrag sind die vom Kunden akzeptierten Geschäftsbedingungen (Business Terms) von eustella oder, sofern die Parteien eine solche geschlossen haben, eine individuelle schriftliche Vereinbarung (z. B. ein Bestellformular oder eine Unternehmensvereinbarung), die den Geschäftsbedingungen im Falle eines Widerspruchs vorgeht. Dieser AVV regelt die Verarbeitung personenbezogener Daten, die eustella im Auftrag des Kunden im Zusammenhang mit den Diensten durchführt. Er wird gemäß Art. 28 Abs. 3 der Verordnung (EU) 2016/679 („DSGVO") geschlossen. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) Auftragsverarbeiter AI Newsrooms Technology GmbH Karl-Farkas-Gasse 22, 1030 Wien, Österreich Firmenbuchnummer: FN 637289s, Handelsgericht Wien UID-Nummer: ATU81169478 Kontakt für Datenschutzangelegenheiten: privacy@eustella.com (nachfolgend „eustella" oder der „Auftragsverarbeiter") Verantwortlicher Vollständige Firmenbezeichnung des KundenAnschriftFirmenbuch- / UID-NummerKontakt für Datenschutzangelegenheiten (nachfolgend der „Kunde" oder der „Verantwortliche") eustella und der Kunde sind jeweils eine „Partei" und gemeinsam die „Parteien". ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) 2.1 Im Rahmen der Erbringung der Dienste nach dem Hauptvertrag verarbeitet eustella personenbezogene Daten im Auftrag und auf dokumentierte Weisung des Kunden. In Bezug auf diese Verarbeitung ist der Kunde Verantwortlicher (bzw., soweit der Kunde selbst als Auftragsverarbeiter für einen Dritten handelt, dessen Auftragsverarbeiter) und eustella Auftragsverarbeiter im Sinne des Art. 4 Nr. 8 DSGVO. 2.2 Jede Partei ist für die Einhaltung ihrer eigenen Pflichten nach dem anwendbaren Datenschutzrecht verantwortlich. Der Kunde ist für die Rechtmäßigkeit der von ihm angewiesenen Verarbeitung verantwortlich, einschließlich des Vorliegens einer gültigen Rechtsgrundlage für die von ihm an die Dienste übermittelten personenbezogenen Daten. 2.3 Gegenstand, Dauer, Art und Zweck der Verarbeitung, die Arten personenbezogener Daten und die Kategorien betroffener Personen sind in Anlage 1 festgelegt. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) Dieser AVV wird mit dem Inkrafttreten des Hauptvertrags wirksam und bleibt in Kraft, solange eustella personenbezogene Daten im Auftrag des Kunden nach dem Hauptvertrag verarbeitet. Pflichten, die ihrem Wesen nach über die Beendigung hinaus fortbestehen — insbesondere die Abschnitte 11 (Löschung und Rückgabe) und 13 (Vertraulichkeit) — gelten über die Beendigung dieses AVV hinaus fort. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) 4.1 eustella verarbeitet personenbezogene Daten ausschließlich auf dokumentierte Weisung des Kunden, auch in Bezug auf die Übermittlung personenbezogener Daten an ein Drittland, es sei denn, eustella ist nach dem Unionsrecht oder dem Recht eines Mitgliedstaats, dem eustella unterliegt, hierzu verpflichtet. In einem solchen Fall teilt eustella dem Kunden diese rechtliche Anforderung vor der Verarbeitung mit, sofern das betreffende Recht eine solche Mitteilung nicht wegen eines wichtigen öffentlichen Interesses verbietet. 4.2 Der Hauptvertrag, dieser AVV (einschließlich seiner Anlagen) sowie die dokumentierte Nutzung der Dienste durch den Kunden stellen die vollständige und abschließende Weisung des Kunden dar. Zusätzliche oder abweichende Weisungen bedürfen der Schriftform und können, sofern sie Änderungen der Dienste erfordern, eine Anpassung der Entgelte nach sich ziehen. 4.3 eustella informiert den Kunden unverzüglich, falls eine Weisung nach Auffassung von eustella gegen die DSGVO oder andere anwendbare Datenschutzbestimmungen verstößt. eustella ist berechtigt, die Ausführung einer solchen Weisung bis zu deren Bestätigung oder Änderung auszusetzen. 4.4 Keine Nutzung zu eigenen Zwecken; kein KI-Training. eustella nutzt personenbezogene Daten des Kunden nicht für eigene Zwecke. Hiervon ausgenommen sind die Gewährleistung der Sicherheit, Integrität und Verfügbarkeit der Dienste, die Abwehr von Missbrauch und Betrug, die Erfüllung gesetzlicher Pflichten sowie die Erstellung aggregierter, nicht personenbezogener Statistiken zur Verbesserung der Dienste. eustella verwendet Kundeninhalte nicht, um KI-Modelle zu trainieren — weder eigene noch solche Dritter. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) eustella gewährleistet, dass sich die zur Verarbeitung der personenbezogenen Daten befugten Personen zur Vertraulichkeit verpflichtet haben oder einer angemessenen gesetzlichen Verschwiegenheitspflicht unterliegen. Der Zugriff auf personenbezogene Daten des Kunden ist auf das Personal beschränkt, das diese zur Erbringung der Dienste benötigt; kein Mitarbeiter hat standardmäßig Zugriff auf den Inhalt von Konversationen oder Kundeneingaben, und ein etwaiger Zugriff ist auf hierfür benanntes Personal in bestimmten, dokumentierten Fällen beschränkt. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) 6.1 eustella trifft geeignete technische und organisatorische Maßnahmen gemäß Art. 32 DSGVO, um ein dem Risiko angemessenes Schutzniveau zu gewährleisten. Die AI Newsrooms Technology GmbH ist nach ISO/IEC 27001:2022 zertifiziert (Zertifikat-Nr. IS-IA-2026-04-15-01). Die zum Zeitpunkt des Inkrafttretens geltenden Maßnahmen sind in Anlage 2 beschrieben. 6.2 eustella kann die Maßnahmen im Laufe der Zeit aktualisieren, sofern das Schutzniveau dadurch nicht wesentlich verringert wird. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) 7.1 Der Kunde erteilt eustella die allgemeine schriftliche Genehmigung zur Beauftragung von Unterauftragsverarbeitern. eustella führt unter eustella.com/services/privacy/sub-processors eine aktuelle Liste der Unterauftragsverarbeiter, die Bestandteil der Anlage 3 ist und die maßgebliche, jeweils aktuelle Liste darstellt. 7.2 eustella erlegt jedem Unterauftragsverarbeiter durch Vertrag Datenschutzpflichten auf, die nicht weniger schützend sind als die in diesem AVV festgelegten, insbesondere die Pflicht, hinreichende Garantien für die Umsetzung geeigneter technischer und organisatorischer Maßnahmen zu bieten. eustella haftet gegenüber dem Kunden für die Erfüllung der Pflichten der jeweiligen Unterauftragsverarbeiter. 7.3 eustella informiert den Kunden über jede beabsichtigte Hinzuziehung oder Ersetzung eines Unterauftragsverarbeiters und gibt dem Kunden die Möglichkeit zum Widerspruch. Der Kunde kann innerhalb von 30 Tagen nach der Information aus berechtigten datenschutzrechtlichen Gründen widersprechen. Können die Parteien den Widerspruch nicht ausräumen, kann der Kunde die Dienste als alleiniges Rechtsmittel kündigen. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) 8.1 EU-only-Verarbeitung. Die Infrastruktur von eustella — die Server, die die Dienste betreiben, Kundendaten speichern und die KI-Inferenz durchführen — wird ausschließlich bei in der Europäischen Union ansässigen Anbietern betrieben. Sämtliche im Rahmen dieses AVV verarbeiteten personenbezogenen Daten werden innerhalb der EU/des EWR verarbeitet und gespeichert. 8.2 eustella übermittelt personenbezogene Daten des Kunden nicht in ein Drittland außerhalb der EU/des EWR und verarbeitet sie nicht dort; eustella beauftragt auch keinen Unterauftragsverarbeiter, der personenbezogene Daten des Kunden außerhalb der EU/des EWR verarbeitet. Alle Unterauftragsverarbeiter sind in der EU/im EWR ansässige Anbieter (siehe Anlage 3). Sollte eustella jemals beabsichtigen, eine solche Übermittlung einzuführen, würde eustella den Kunden zuvor gemäß Abschnitt 7 informieren und einen geeigneten Übermittlungsmechanismus nach Kapitel V DSGVO einrichten; andernfalls verbleibt die gesamte Verarbeitung innerhalb der EU/des EWR. 8.3 Verbundene Drittdienste. Sofern der Kunde oder seine Endnutzer sich dafür entscheiden, Drittdienste mit den Diensten zu verbinden (z. B. Google Kalender, Google Drive oder andere Kalender-, Cloud-Speicher- oder Produktivitätsdienste), erfolgt jede Übermittlung personenbezogener Daten an einen solchen Anbieter bzw. jede Verarbeitung durch ihn im Rahmen des eigenen Verantwortlichenverhältnisses des Kunden zu diesem Anbieter und nach dessen eigenen Bedingungen. Solche Anbieter sind keine Unterauftragsverarbeiter von eustella, sind nicht in Anlage 3 aufgeführt, und ein daraus resultierender Datenfluss — einschließlich einer etwaigen Übermittlung in ein Drittland außerhalb der EU/des EWR — stellt weder eine Unterauftragsverarbeitung noch eine durch eustella durchgeführte Übermittlung dar. Die EU-only-Zusagen in diesem Abschnitt 8 gelten für die eigene Verarbeitungsinfrastruktur von eustella und ihre Unterauftragsverarbeiter; sie regeln nicht die eigenständige Nutzung verbundener Drittdienste durch den Kunden, für die der Kunde verantwortlich ist (einschließlich der Beurteilung und, soweit einschlägig, der Schaffung einer gültigen Grundlage für eine etwaige Drittlandübermittlung nach Kapitel V DSGVO). ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) Unter Berücksichtigung der Art der Verarbeitung unterstützt eustella den Kunden mit geeigneten technischen und organisatorischen Maßnahmen, soweit dies möglich ist, bei der Erfüllung seiner Pflicht zur Beantwortung von Anträgen auf Wahrnehmung der in Kapitel III DSGVO (Art. 12–23) genannten Rechte der betroffenen Personen. Über die zur Erfüllung der gesetzlichen Pflichten erforderliche, zumutbare Mitwirkung hinausgehende Unterstützungsleistungen vergütet der Kunde nach dem vereinbarten Aufwand. Wendet sich eine betroffene Person unmittelbar an eustella in Bezug auf personenbezogene Daten, die im Auftrag des Kunden verarbeitet werden, leitet eustella den Antrag unverzüglich an den Kunden weiter und antwortet nicht selbst, es sei denn, der Kunde weist dies an oder eustella ist gesetzlich dazu verpflichtet. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) Unter Berücksichtigung der Art der Verarbeitung und der eustella zur Verfügung stehenden Informationen unterstützt eustella den Kunden bei der Einhaltung der Pflichten nach Art. 32 bis 36 DSGVO, einschließlich der Sicherheit der Verarbeitung, der Meldung von Verletzungen des Schutzes personenbezogener Daten, der Datenschutz-Folgenabschätzung und der vorherigen Konsultation der Aufsichtsbehörde. Über die zur Erfüllung der gesetzlichen Pflichten erforderliche, zumutbare Mitwirkung hinausgehende Unterstützungsleistungen vergütet der Kunde nach dem vereinbarten Aufwand. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) eustella benachrichtigt den Kunden unverzüglich, nachdem ihr eine Verletzung des Schutzes personenbezogener Daten des Kunden mit hinreichender Sicherheit bekannt geworden ist. Die Benachrichtigung beschreibt, soweit bekannt, die Art der Verletzung, die wahrscheinlichen Folgen, die Kategorien und die ungefähre Zahl der betroffenen Personen und Datensätze sowie die ergriffenen oder vorgeschlagenen Maßnahmen zu deren Behebung. Liegen zum Zeitpunkt der Erstmeldung noch nicht alle Informationen vor, kann eustella diese schrittweise nachreichen. Die Meldung stellt kein Anerkenntnis eines Verschuldens oder einer Haftung dar. eustella benachrichtigt Aufsichtsbehörden oder betroffene Personen nicht im Namen des Kunden, es sei denn, sie wird hierzu angewiesen. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) 12.1 eustella stellt dem Kunden alle Informationen zur Verfügung, die zum Nachweis der Einhaltung der in Art. 28 DSGVO und in diesem AVV festgelegten Pflichten vernünftigerweise erforderlich sind. 12.2 Das Recht des Kunden, die Einhaltung durch eustella zu überprüfen, wird vorrangig durch Unterlagen erfüllt und kann höchstens einmal pro Kalenderjahr ausgeübt werden; dies gilt nicht, soweit eine zuständige Aufsichtsbehörde eine weitergehende Prüfung anordnet. Auf schriftliche Anfrage stellt eustella einschlägige Compliance-Unterlagen bereit — insbesondere die ISO/IEC 27001:2022-Zertifizierung der AI Newsrooms Technology GmbH (Zertifikat-Nr. IS-IA-2026-04-15-01), Sicherheitsbeschreibungen sowie, soweit vorhanden, unabhängige Prüf- oder Bewertungsberichte Dritter. Vor-Ort-Prüfungen und physische Inspektionen der Räumlichkeiten, Systeme oder Einrichtungen von eustella sind nur zulässig, soweit die bereitgestellten Unterlagen zum Nachweis der Einhaltung nach Art. 28 DSGVO nachweislich nicht ausreichen oder eine zuständige Aufsichtsbehörde dies anordnet. Eine solche Prüfung erfolgt nach angemessener Vorankündigung von mindestens 30 Tagen, zu üblichen Geschäftszeiten, unter Wahrung der Vertraulichkeit sowie der Sicherheit und der personenbezogenen Daten anderer Kunden, und auf Kosten des Kunden. Die Verpflichtung von eustella nach diesem Abschnitt steht unter dem Vorbehalt der Vertraulichkeit und beschränkt sich im Übrigen auf die eustella bereits vorliegenden Unterlagen, sofern nicht eine zuständige Aufsichtsbehörde etwas anderes anordnet. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) Nach Wahl des Kunden löscht eustella nach Abschluss der Erbringung der Dienste alle personenbezogenen Daten des Kunden oder gibt sie zurück und löscht vorhandene Kopien, sofern nicht nach dem Unionsrecht oder dem Recht eines Mitgliedstaats eine Verpflichtung zur Speicherung der personenbezogenen Daten besteht. Gesetzliche Aufbewahrungspflichten (etwa nach österreichischem Buchhaltungsrecht) sowie eine kurzfristige Aufbewahrung in routinemäßigen Sicherungskopien, die im ordentlichen Turnus gelöscht werden, bleiben unberührt. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) 14.1 Die Haftung richtet sich nach Art. 82 DSGVO und nach den Haftungsbestimmungen des Hauptvertrags. 14.2 Dieser AVV unterliegt österreichischem Recht unter Ausschluss der Kollisionsnormen und des UN-Kaufrechts. Ausschließlicher Gerichtsstand ist, soweit gesetzlich zulässig, Wien, Österreich. ## Auftragsverarbeitungsvertrag › eustella Auftragsverarbeitungsvertrag (AVV) 15.1 Im Falle eines Widerspruchs zwischen diesem AVV und dem Hauptvertrag hinsichtlich der Verarbeitung personenbezogener Daten geht dieser AVV vor. 15.2 Sollte eine Bestimmung dieses AVV unwirksam sein oder werden, so wird die Wirksamkeit der übrigen Bestimmungen davon nicht berührt. ## Auftragsverarbeitungsvertrag Gegenstand: Verarbeitung personenbezogener Daten durch eustella zur Erbringung der Dienste für den Kunden nach dem Hauptvertrag. Dauer: Für die Laufzeit des Hauptvertrags und bis zur Löschung oder Rückgabe der personenbezogenen Daten gemäß Abschnitt 13. Art und Zweck: Hosting, Speicherung, Übertragung, KI-Inferenz (Erzeugung von Modellausgaben aus Kundeneingaben) sowie damit verbundene Verarbeitungsvorgänge, die zur Erbringung der Dienste erforderlich sind, einschließlich — je nach genutzten Diensten: Chat-/Agent-Antworten, Websuche, Texterkennung (OCR), Audiotranskription und Bildgenerierung. Arten personenbezogener Daten: Alle personenbezogenen Daten, die in den Eingaben, Prompts, Dateien, Dokumenten und sonstigen Inhalten enthalten sind, die der Kunde oder seine Endnutzer an die Dienste übermitteln, sowie in den daraus erzeugten Ausgaben; Konto- und Authentifizierungsdaten; sowie technische und Nutzungsdaten (etwa IP-Adresse, Geräte- und Browserinformationen sowie Ereignisprotokolle). Die konkreten Kategorien werden durch den Kunden im Rahmen seiner Nutzung der Dienste bestimmt. Besondere Kategorien personenbezogener Daten (Art. 9 DSGVO): Nicht vorgesehen. Der Kunde ist dafür verantwortlich, keine Daten besonderer Kategorien zu übermitteln, es sei denn, er hat eine gültige Rechtsgrundlage sichergestellt und eustella informiert, wo zusätzliche Schutzmaßnahmen erforderlich sind. Kategorien betroffener Personen: Endnutzer des Kunden sowie alle Personen, deren personenbezogene Daten in den vom Kunden an die Dienste übermittelten Inhalten enthalten sind (dies können Kunden, Beschäftigte, Kontakte oder sonstige Dritte des Kunden sein). ## Auftragsverarbeitungsvertrag Zertifizierung: Die AI Newsrooms Technology GmbH ist nach ISO/IEC 27001:2022 zertifiziert (Zertifikat-Nr. IS-IA-2026-04-15-01); die Zertifizierung umfasst ihr Informationssicherheits-Managementsystem. Es werden unabhängige Black-Box-Penetrationstests der Produktivplattform durchgeführt, einschließlich Nachverfolgung und Behebung festgestellter Schwachstellen.Verschlüsselung: Personenbezogene Daten werden bei der Übertragung (TLS 1.2 oder höher, aktuelle Cipher Suites) und im Ruhezustand (AES-256 oder gleichwertige branchenübliche Verfahren) verschlüsselt. Vertrauliche Konfigurationswerte, API-Schlüssel und sonstige Geheimnisse werden ausschließlich in dafür vorgesehenen Secret-Management-/Key-Management-Systemen vorgehalten.Zugriffskontrolle und Authentifizierung: Rollenbasierte Zugriffskontrollen und Zuweisung nach dem Least-Privilege-/Need-to-know-Prinzip mit regelmäßiger Überprüfung; verpflichtende Multi-Faktor-Authentifizierung (2FA) für alle privilegierten und administrativen Zugriffe (z. B. Cloud-Konsole, CI/CD-Pipelines, Administration); strenge Passwort- und Zugangsrichtlinien; kein standardmäßiger Mitarbeiterzugriff auf Kundeninhalte, Zugriff auf benanntes Personal in bestimmten, dokumentierten Fällen beschränkt.EU-Datenresidenz: Sämtliche Verarbeitung und Speicherung erfolgen innerhalb der EU/des EWR; keine internationalen Übermittlungen (siehe Abschnitt 8). Alle Unterauftragsverarbeiter sind in der EU/im EWR ansässige Anbieter.Netzwerk- und Infrastruktursicherheit: Betrieb der Plattforminfrastruktur in logisch isolierten Netzwerken (VPCs, Subnetze); Segmentierung nach Umgebung; Firewall-Regeln und Network Policies, die den Zugriff auf notwendige Ports und Protokolle beschränken; Nutzung einer zugriffsbeschränkten, zentral verwalteten Container Registry; regelmäßiges Patch- und Schwachstellenmanagement der Basis-Images und Abhängigkeiten.Trennung von Umgebungen und Mandanten: Klare Trennung von Entwicklungs-, Staging- und Produktionsumgebungen; produktive Kundendaten werden nicht für Entwicklung oder Tests verwendet; logische Mandanten-/Kundentrennung auf Datenbank- und/oder Anwendungsebene (z. B. separate Datenbanken/Schemata, Mandanten-IDs).Integrität: Zentrale Protokollierung relevanter Zugriffe und administrativer Handlungen; Schutz vor unbefugter Veränderung.Verfügbarkeit und Belastbarkeit: Redundante, mehrregionale europäische Infrastruktur; routinemäßige Sicherungen, die getrennt innerhalb der EU aufbewahrt werden; regelmäßige Wiederherstellungstests anhand definierter Recovery Time und Recovery Point Objectives (RTO/RPO); Maßnahmen zur zeitnahen Wiederherstellung der Verfügbarkeit und des Zugangs zu personenbezogenen Daten nach einem Vorfall.Überwachung und Reaktion auf Vorfälle: Überwachung von Infrastruktur und Diensten mit Alarmierung bei Anomalien; etablierte interne Prozesse zur Bewertung und Behandlung von Sicherheitsvorfällen einschließlich Eskalationswegen und Dokumentation; definiertes Verfahren zur Meldung von Datenpannen an den Verantwortlichen (Mindestinhalt und Fristen gemäß Abschnitt 11).Kein Modelltraining mit Kundendaten: Kundeninhalte werden nicht zum Training, zur Feinabstimmung (Fine-Tuning) oder zur Verbesserung von KI-Modellen verwendet — weder eigene Modelle von eustella noch solche Dritter.Steuerung von Unterauftragsverarbeitern: Allen Unterauftragsverarbeitern werden vertragliche Datenschutzpflichten auferlegt, die nicht weniger schützend sind als dieser AVV; geführte Unterauftragsverarbeiter-Liste.Evaluierung: Regelmäßige interne Überprüfung (Sicherheits- und Architektur-Reviews) zur Bewertung und gegebenenfalls Verbesserung der Wirksamkeit dieser Maßnahmen. ## Auftragsverarbeitungsvertrag Die maßgebliche, jeweils aktuelle Liste der Unterauftragsverarbeiter wird unter eustella.com/services/privacy/sub-processors veröffentlicht und gepflegt. Die nachstehende Liste gibt die zum Zeitpunkt des Inkrafttretens beauftragten Unterauftragsverarbeiter wieder. Alle für die Geschäftskundenverarbeitung (B2B) beauftragten Unterauftragsverarbeiter sind in der EU/im EWR ansässig und verarbeiten personenbezogene Daten des Kunden ausschließlich innerhalb der EU/des EWR. Welche Unterauftragsverarbeiter Anwendung finden, hängt von den vom Kunden genutzten Diensten ab. UnterauftragsverarbeiterRechtsträger / StandortZweckÜbermittlungsgarantieIONOSIONOS SE — Deutschland (EU)Kern-Cloud-Infrastruktur; Server und DatenspeicherungEU — nicht erforderlichScalewayScaleway SAS — Frankreich (EU)Ergänzende Cloud-Rechenleistung und KI-WorkloadsEU — nicht erforderlichMistral AIMistral AI — Frankreich (EU)OCR und AudiotranskriptionEU — nicht erforderlichVerdaVerda Cloud Oy — Finnland (EU)GPU-Rechenleistung für KI-ModellinferenzEU — nicht erforderlichLinkupLinkup Technologies SAS — Frankreich (EU)Echtzeit-WebsucheEU — nicht erforderlich --- # eustella — Data Processing Agreement URL: https://eustella.com/services/dpa Topic: services Last updated: 2026-07-06 ## Data Processing Agreement Last updated: 06.07.2026 · Version 1.2.0 Deutsche Fassung (Auftragsverarbeitungsvertrag): eustella.com/de/services/dpa. In case of conflict, the language version executed by the parties prevails. This Data Processing Agreement ("DPA") is intended solely for eustella's business (B2B) customers — organisations that use eustella's services for professional or commercial purposes. It does not apply to consumers using eustella's mobile applications or consumer features. This DPA forms part of, and is incorporated by reference into, the agreement between the Customer and eustella for the use of eustella's services (the "Main Agreement"). The Main Agreement is eustella's Business Terms as accepted by the Customer or, where the Parties have concluded one, an individual written agreement (for example, an order form or enterprise agreement), which prevails over the Business Terms to the extent of any conflict. This DPA governs the processing of personal data carried out by eustella on behalf of the Customer in connection with the Services. It is concluded pursuant to Article 28(3) of Regulation (EU) 2016/679 ("GDPR"). ## Data Processing Agreement › eustella Data Processing Agreement Processor AI Newsrooms Technology GmbH Karl-Farkas-Gasse 22, 1030 Vienna, Austria Company register: FN 637289s, Commercial Court of Vienna VAT ID: ATU81169478 Contact for data protection matters: privacy@eustella.com (hereinafter "eustella" or the "Processor") Controller Customer legal nameAddressCompany register / VAT IDContact for data protection matters (hereinafter the "Customer" or the "Controller") eustella and the Customer are each a "Party" and together the "Parties". ## Data Processing Agreement › eustella Data Processing Agreement 2.1 In the course of providing the Services under the Main Agreement, eustella processes personal data on behalf of and on the documented instructions of the Customer. With respect to such processing, the Customer acts as controller (or, where the Customer itself acts as a processor for a third party, as that third party's processor) and eustella acts as processor within the meaning of Article 4(8) GDPR. 2.2 Each Party is responsible for complying with its own obligations under applicable data protection law. The Customer is responsible for the lawfulness of the processing it instructs, including for having a valid legal basis for the personal data it submits to the Services. 2.3 The subject matter, duration, nature and purpose of the processing, the types of personal data and the categories of data subjects are set out in Annex 1. ## Data Processing Agreement › eustella Data Processing Agreement This DPA takes effect on the effective date of the Main Agreement and remains in force for as long as eustella processes personal data on behalf of the Customer under the Main Agreement. Obligations that by their nature survive termination — in particular Sections 11 (deletion and return) and 13 (confidentiality) — survive termination of this DPA. ## Data Processing Agreement › eustella Data Processing Agreement 4.1 eustella processes personal data only on the documented instructions of the Customer, including with regard to transfers of personal data to a third country, unless required to do so by Union or Member State law to which eustella is subject. In such a case, eustella informs the Customer of that legal requirement before processing, unless that law prohibits such information on important grounds of public interest. 4.2 The Main Agreement, this DPA (including its Annexes), and the Customer's documented use of the Services constitute the Customer's complete and final instructions. Additional or alternative instructions must be agreed in writing and may be subject to adjustment of fees if they require changes to the Services. 4.3 eustella informs the Customer without undue delay if, in its opinion, an instruction infringes the GDPR or other applicable data protection provisions. eustella is entitled to suspend execution of such an instruction until it is confirmed or amended. 4.4 No use for own purposes; no model training. eustella does not use Customer personal data for its own purposes. This does not apply to ensuring the security, integrity and availability of the services, the prevention of abuse and fraud, compliance with legal obligations, and the creation of aggregated, non-personal statistics to improve the services. eustella does not use Customer content to train AI models — neither its own nor those of any third party. ## Data Processing Agreement › eustella Data Processing Agreement eustella ensures that persons authorised to process the personal data have committed themselves to confidentiality or are under an appropriate statutory obligation of confidentiality. Access to Customer personal data is limited to personnel who need it to provide the Services; no employee has default access to the content of conversations or Customer inputs, and any such access is restricted to designated staff in specific, documented circumstances. ## Data Processing Agreement › eustella Data Processing Agreement 6.1 eustella implements appropriate technical and organisational measures pursuant to Article 32 GDPR to ensure a level of security appropriate to the risk. AI Newsrooms Technology GmbH is certified to ISO/IEC 27001:2022 (Certificate No. IS-IA-2026-04-15-01). The measures in force at the effective date are described in Annex 2. 6.2 eustella may update the measures over time provided that the level of security is not materially reduced. ## Data Processing Agreement › eustella Data Processing Agreement 7.1 The Customer grants eustella general written authorisation to engage sub-processors. eustella maintains a current list of sub-processors at eustella.com/services/privacy/sub-processors, which forms part of Annex 3 and is the authoritative, up-to-date list. 7.2 eustella imposes on each sub-processor, by way of a contract, data protection obligations that are no less protective than those set out in this DPA, in particular the obligation to provide sufficient guarantees to implement appropriate technical and organisational measures. eustella remains liable to the Customer for the performance of each sub-processor's obligations. 7.3 eustella informs the Customer of any intended addition or replacement of a sub-processor, giving the Customer the opportunity to object. The Customer may object on reasonable data-protection grounds within 30 days of being informed. If the Parties cannot resolve the objection, the Customer may terminate the Services as its sole remedy. ## Data Processing Agreement › eustella Data Processing Agreement 8.1 EU-only processing. eustella's infrastructure — the servers that run the Services, store Customer data, and perform AI inference — is hosted exclusively with providers established in the European Union. All personal data processed under this DPA is processed and stored within the EU/EEA. 8.2 eustella does not transfer Customer personal data to, or process it in, any third country outside the EU/EEA, and does not engage any sub-processor that processes Customer personal data outside the EU/EEA. All sub-processors are EU/EEA-established providers (see Annex 3). Were eustella ever to intend to introduce any such transfer, it would first notify the Customer in accordance with Section 7 and put in place an appropriate transfer mechanism under Chapter V GDPR; absent that, all processing remains within the EU/EEA. 8.3 Connected third-party services. Where the Customer or its end users choose to connect third-party services to the Services (for example, Google Calendar, Google Drive or other calendars, cloud storage or productivity tools), any transfer of personal data to, or processing by, such a provider takes place under the Customer's own controller relationship with that provider and under that provider's own terms. Such providers are not eustella sub-processors, are not listed in Annex 3, and any resulting data flow — including any transfer to a third country outside the EU/EEA — is neither sub-processing nor a transfer carried out by eustella. The EU-only commitments in this Section 8 apply to eustella's own processing infrastructure and its sub-processors; they do not govern the Customer's independent use of connected third-party services, for which the Customer is responsible (including for assessing and, where applicable, providing a valid basis for any third-country transfer under Chapter V GDPR). ## Data Processing Agreement › eustella Data Processing Agreement Taking into account the nature of the processing, eustella assists the Customer by appropriate technical and organisational measures, insofar as this is possible, in fulfilling the Customer's obligation to respond to requests for exercising the data subject's rights under Chapter III GDPR (Articles 12–23). Support services exceeding the reasonable cooperation required to fulfil statutory obligations shall be remunerated by the customer on a time-and-materials basis as agreed. If a data subject contacts eustella directly regarding personal data processed on behalf of the Customer, eustella forwards the request to the Customer without undue delay and does not respond itself unless instructed by the Customer or legally required. ## Data Processing Agreement › eustella Data Processing Agreement Taking into account the nature of processing and the information available to eustella, eustella assists the Customer in ensuring compliance with the obligations under Articles 32 to 36 GDPR, including security of processing, notification of personal data breaches, data protection impact assessments, and prior consultation with the supervisory authority. Support services exceeding the reasonable cooperation required to fulfil statutory obligations shall be remunerated by the customer on a time-and-materials basis as agreed. ## Data Processing Agreement › eustella Data Processing Agreement eustella notifies the Customer without undue delay after becoming aware, with reasonable certainty, of a personal data breach affecting Customer personal data. The notification describes, to the extent known, the nature of the breach, the likely consequences, the categories and approximate number of data subjects and records concerned, and the measures taken or proposed to address it. Where not all information is available at the time of the initial notification, eustella may provide it in phases. The notification does not constitute an acknowledgement of fault or liability. eustella does not notify supervisory authorities or data subjects on the Customer's behalf unless instructed to do so. ## Data Processing Agreement › eustella Data Processing Agreement 12.1 eustella makes available to the Customer all information reasonably necessary to demonstrate compliance with the obligations laid down in Article 28 GDPR and this DPA. 12.2 The Customer's right to verify eustella's compliance is satisfied primarily by documentation and may be exercised no more than once per year; this does not apply where a competent supervisory authority orders a further audit. On written request, eustella provides relevant compliance documentation — in particular AI Newsrooms Technology GmbH's ISO/IEC 27001:2022 certification (Certificate No. IS-IA-2026-04-15-01), security descriptions, and independent third-party audit or assessment reports, where available. On-site audits and physical inspections of eustella's premises, systems or facilities are permitted only where the documentation provided is demonstrably insufficient to evidence compliance under Art. 28 GDPR or where a competent supervisory authority so orders. Any such audit shall take place upon reasonable prior notice of at least 30 days, during normal business hours, while preserving confidentiality and the security and personal data of other customers, and at the customer's expense. eustella's obligation under this section is subject to confidentiality and is otherwise limited to the documentation already available to eustella, unless a competent supervisory authority requires otherwise. ## Data Processing Agreement › eustella Data Processing Agreement At the Customer's choice, eustella deletes or returns all Customer personal data after the end of the provision of the Services and deletes existing copies, unless Union or Member State law requires storage of the personal data. Statutory retention obligations (for example under Austrian accounting law) and short-term retention in routine backups, which are deleted on their ordinary cycle, are unaffected. ## Data Processing Agreement › eustella Data Processing Agreement 14.1 Liability is governed by Article 82 GDPR and by the liability provisions of the Main Agreement. 14.2 This DPA is governed by the laws of Austria, excluding its conflict-of-law rules and the UN Convention on Contracts for the International Sale of Goods. The exclusive place of jurisdiction is Vienna, Austria, to the extent permitted by law. ## Data Processing Agreement › eustella Data Processing Agreement 15.1 In the event of a conflict between this DPA and the Main Agreement regarding the processing of personal data, this DPA prevails. 15.2 Should any provision of this DPA be or become invalid, the validity of the remaining provisions is not affected. ## Data Processing Agreement Subject matter: Processing of personal data by eustella in order to provide the Services to the Customer under the Main Agreement. Duration: For the term of the Main Agreement and until deletion or return of the personal data in accordance with Section 13. Nature and purpose: Hosting, storage, transmission, AI inference (generation of model outputs from Customer inputs), and related processing operations necessary to provide the Services, including, depending on the Services used: chat/agent completions, web search, optical character recognition (OCR), audio transcription, and image generation. Types of personal data: Any personal data contained in the inputs, prompts, files, documents and other content that the Customer or its end users submit to the Services, and in the outputs generated therefrom; account and authentication data; and technical and usage data (such as IP address, device and browser information, and event logs). The specific categories are determined by the Customer through its use of the Services. Special categories of personal data (Art. 9 GDPR): Not intended. The Customer is responsible for not submitting special-category data unless it has ensured a valid legal basis and has informed eustella where additional safeguards are required. Categories of data subjects: The Customer's end users and any individuals whose personal data is contained in the content the Customer submits to the Services (which may include the Customer's customers, employees, contacts or other third parties). ## Data Processing Agreement Certification: AI Newsrooms Technology GmbH is certified to ISO/IEC 27001:2022 (Certificate No. IS-IA-2026-04-15-01), covering its information-security management system. Independent black-box penetration testing of the production platform is conducted, with tracking and remediation of identified findings.Encryption: Personal data is encrypted in transit (TLS 1.2 or higher, current cipher suites) and at rest (AES-256 or equivalent industry-standard protocols). Confidential configuration values, API keys and other secrets are held exclusively in dedicated secret-management / key-management systems.Access control and authentication: Role-based access controls and least-privilege/need-to-know assignment, reviewed periodically; mandatory multi-factor authentication (2FA) for all privileged and administrative access (e.g. cloud console, CI/CD pipelines, administration); strict password and access policies; no default employee access to Customer content, with access limited to designated staff in specific, documented circumstances.EU data residency: All processing and storage take place within the EU/EEA; no international transfers (see Section 8). All sub-processors are EU/EEA-established providers.Network and infrastructure security: Operation of platform infrastructure in logically isolated networks (VPCs, subnets); segmentation by environment; firewall rules and network policies restricting access to necessary ports and protocols; use of an access-controlled, centrally managed container registry; regular patch and vulnerability management of base images and dependencies.Separation of environments and tenants: Clear separation of development, staging and production environments; production Customer data is not used for development or testing; logical tenant/customer isolation at the database and/or application layer (e.g. separate databases/schemata, tenant IDs).Integrity: Central logging of relevant access and administrative actions; protection against unauthorised modification.Availability and resilience: Redundant, multi-region European infrastructure; routine backups, retained separately within the EU; regular restore testing against defined Recovery Time and Recovery Point Objectives (RTO/RPO); measures to restore availability and access to personal data in a timely manner after an incident.Monitoring and incident response: Monitoring of infrastructure and services with alerting on anomalies; established internal processes for evaluating and handling security incidents, including escalation paths and documentation; defined breach-notification procedure to the Controller (minimum content and timing per Section 11).No model training on Customer data: Customer content is not used to train, fine-tune or improve AI models — neither eustella's own nor those of any third party.Sub-processor governance: Contractual data-protection obligations imposed on all sub-processors that are no less protective than this DPA; maintained sub-processor list.Evaluation: Regular internal review (security and architecture reviews) to assess and, where appropriate, improve the effectiveness of these measures. ## Data Processing Agreement The authoritative, current list of sub-processors is published and maintained at eustella.com/services/privacy/sub-processors. The list below reflects the sub-processors engaged as at the effective date. All sub-processors engaged for business (B2B) processing are established in the EU/EEA and process Customer personal data exclusively within the EU/EEA. Which sub-processors apply depends on the Services the Customer uses. Sub-processorLegal entity / locationPurposeTransfer safeguardIONOSIONOS SE — Germany (EU)Core cloud infrastructure; servers and data storageEU — none requiredScalewayScaleway SAS — France (EU)Supplementary cloud compute and AI workloadsEU — none requiredMistral AIMistral AI — France (EU)OCR and audio transcriptionEU — none requiredVerdaVerda Cloud Oy — Finland (EU)GPU compute for AI model inferenceEU — none requiredLinkupLinkup Technologies SAS — France (EU)Real-time web searchEU — none required --- # eustella — Datenschutz URL: https://eustella.com/services/privacy/de Topic: services ## Datenschutz Stand: 22.06.2026 · Version 1.0.1 English version (Privacy Policy): eustella.com/services/privacy. Die englische Fassung ist die Originalfassung. Diese Datenschutzerklärung erläutert, wie wir Ihre personenbezogenen Daten erheben, verwenden und schützen, wenn Sie eustella nutzen. Wir haben sie klar und ehrlich verfasst — sollte etwas unklar sein, kontaktieren Sie uns bitte, und wir erklären es Ihnen gerne. Diese Datenschutzerklärung ist Bestandteil unserer Beziehung zu Ihnen, ist jedoch von unseren Nutzungsbedingungen getrennt. Sie gilt für Ihre Nutzung der mobilen Anwendungen, Weboberflächen, KI-Agenten-Funktionen und aller damit verbundenen Dienste von eustella („Dienste"). Wenn Sie die Datenschutzerklärung für unsere Website eustella.com suchen, besuchen Sie bitte eustella.com/privacy. ## Datenschutz › eustella Datenschutzerklärung eustella wird von der AI Newsrooms Technology GmbH entwickelt, einem europäischen KI-Unternehmen mit Sitz in Wien, Österreich. Wir sind der Verantwortliche für Ihre personenbezogenen Daten, das heißt, wir entscheiden, wie und warum Ihre Daten verarbeitet werden. AI Newsrooms Technology GmbH Karl-Farkas-Gasse 22, 1030 Wien, Österreich Firmenbuchnummer: FN 637289s, Handelsgericht Wien E-Mail: privacy@eustella.com Wenn Sie Fragen dazu haben, wie wir mit Ihren Daten umgehen, kontaktieren Sie uns bitte unter privacy@eustella.com. Wir bemühen uns, innerhalb von 14 Tagen zu antworten. ## Datenschutz › eustella Datenschutzerklärung Wir erheben je nach Art Ihrer Nutzung von eustella unterschiedliche Arten von Daten. Hier ein Überblick. Daten, die Sie uns direkt geben Wenn Sie ein Konto erstellen, erheben wir Ihren Namen, Ihre E-Mail-Adresse und Ihr Passwort. Wenn Sie sich über Google oder Apple anmelden, erhalten wir die Profilinformationen, die diese Dienste mit uns teilen (in der Regel Ihren Namen und Ihre E-Mail-Adresse). Wir erhalten oder speichern Ihr Google- oder Apple-Passwort nicht. Wenn Sie eustella nutzen, erheben wir den Inhalt Ihrer Unterhaltungen — Ihre Eingaben (was Sie schreiben oder hochladen) und die Ausgaben von eustella (was eustella für Sie erzeugt). Wenn Sie Dateien, Bilder oder Dokumente hochladen, verarbeiten wir auch diese. Wenn Sie Drittdienste wie Google Calendar oder Google Drive verbinden, greifen wir in Ihrem Auftrag auf Daten aus diesen Diensten zu und verarbeiten sie, und zwar nur in dem Umfang, der zur Bereitstellung der von Ihnen aktivierten Funktionen erforderlich ist. Wir greifen zu keinem anderen Zweck auf Daten aus Ihren verbundenen Diensten zu. Daten, die wir automatisch erheben Wenn Sie unsere Dienste nutzen, erheben wir automatisch bestimmte technische Daten und Nutzungsdaten. Dazu gehören Ihre IP-Adresse, Ihr Gerätetyp, Ihr Betriebssystem, Ihr Browsertyp, Ihre Spracheinstellungen und Ihr ungefährer Standort (abgeleitet aus Ihrer IP-Adresse auf Ebene von Land oder Stadt — wir erheben keine genauen GPS-Standortdaten). Wir erheben außerdem Informationen darüber, wie Sie eustella nutzen, etwa welche Funktionen Sie verwenden, wann Sie sie verwenden und wie Sie mit der App interagieren. Diese Daten werden über PostHog erhoben, unseren Analyseanbieter, der in der Europäischen Union (Frankfurt, Deutschland) gehostet wird. Daten, die eustella über Sie lernt eustella ist ein personalisierter KI-Assistent. Eine seiner Kernfunktionen besteht darin, im Laufe der Zeit Ihre Präferenzen zu lernen, um Ihnen nützlichere und relevantere Antworten geben zu können. Das bedeutet, dass eustella sich Dinge merken kann wie Ihre bevorzugte Sprache, Ihren Kommunikationsstil, die Themen, die Sie interessieren, oder die Art, wie Sie Informationen präsentiert bekommen möchten. Wie genau das funktioniert und welche Kontrolle Sie haben, erläutern wir in Abschnitt 5. ## Datenschutz › eustella Datenschutzerklärung Die DSGVO verlangt von uns, für jede Art der Verwendung Ihrer personenbezogenen Daten einen konkreten rechtlichen Grund zu haben. Hier ein Überblick. Bereitstellung und Betrieb von eustella — Wenn wir Ihre Unterhaltungen verarbeiten, Antworten erzeugen, Verbindungen zu Ihren Drittdiensten herstellen, Ihre Präferenzen speichern (wie in Abschnitt 5 beschrieben) und eustella allgemein für Sie funktionsfähig machen, tun wir dies, weil es zur Erfüllung unseres Vertrags mit Ihnen erforderlich ist (Art. 6 Abs. 1 lit. b DSGVO). Sie haben uns gebeten, diese Dienste bereitzustellen, und wir müssen Ihre Daten verarbeiten, um dies zu tun. Sicherheit von eustella — Wir betreiben eigene Systeme zur Inhaltsmoderation und Sicherheit, um Missbrauch zu erkennen, unsere Nutzungsrichtlinien durchzusetzen und unsere Nutzer sowie die Öffentlichkeit zu schützen. Wir tun dies auf Grundlage unseres berechtigten Interesses an der Aufrechterhaltung eines sicheren und rechtmäßigen Dienstes (Art. 6 Abs. 1 lit. f DSGVO) und, soweit erforderlich, zur Erfüllung unserer rechtlichen Verpflichtungen nach geltenden Gesetzen und Vorschriften, einschließlich des Digital Services Act (Art. 6 Abs. 1 lit. c DSGVO). Verbesserung unserer Dienste — Wir verwenden aggregierte und anonymisierte Nutzungsdaten (etwa welche Funktionen beliebt sind, wo Nutzer auf Fehler stoßen und allgemeine Nutzungsmuster), um eustella zu verbessern. Wir stützen uns dabei auf unser berechtigtes Interesse an der Entwicklung und Verbesserung unseres Dienstes (Art. 6 Abs. 1 lit. f DSGVO). Wir verwenden die Inhalte Ihrer Unterhaltungen nicht, um KI-Modelle zu trainieren oder zu verbessern. Analyse — Wir verwenden PostHog (in der EU gehostet), um zu verstehen, wie Menschen eustella nutzen, damit wir Probleme erkennen und das Erlebnis verbessern können. Für wesentliche Analysen stützen wir uns auf unser berechtigtes Interesse am zuverlässigen Betrieb und an der Verbesserung unseres Dienstes (Art. 6 Abs. 1 lit. f DSGVO). Für nicht wesentliche Analysen stützen wir uns auf Ihre Einwilligung (Art. 6 Abs. 1 lit. a DSGVO), die Sie jederzeit in Ihren Kontoeinstellungen oder durch Kontaktaufnahme mit uns erteilen oder widerrufen können. Der Widerruf der Einwilligung berührt nicht die Rechtmäßigkeit der vor dem Widerruf erfolgten Verarbeitung. Kommunikation mit Ihnen — Wir können Ihnen dienstbezogene Mitteilungen senden, etwa Sicherheitshinweise, Änderungen unserer Bedingungen oder wichtige Aktualisierungen zu Ihrem Konto. Diese sind zur Erfüllung unseres Vertrags mit Ihnen erforderlich (Art. 6 Abs. 1 lit. b DSGVO). Sollten wir Ihnen jemals Marketingmitteilungen senden, tun wir dies nur mit Ihrer ausdrücklichen Einwilligung (Art. 6 Abs. 1 lit. a DSGVO), die Sie jederzeit widerrufen können. Einhaltung des Rechts — Wir verarbeiten Daten, wenn dies gesetzlich vorgeschrieben ist, etwa um auf berechtigte behördliche Anfragen zu reagieren oder um unsere steuer- und buchhaltungsrechtlichen Pflichten nach österreichischem Recht zu erfüllen (Art. 6 Abs. 1 lit. c DSGVO). Schutz unserer Rechte — In seltenen Fällen können wir Daten verarbeiten, um Rechtsansprüche geltend zu machen, auszuüben oder zu verteidigen. Wir stützen uns dabei auf unser berechtigtes Interesse am Schutz unserer Rechtsposition (Art. 6 Abs. 1 lit. f DSGVO). ## Datenschutz › eustella Datenschutzerklärung Manche Dinge sind ebenso wichtig, sie klar auszusprechen. Wir trainieren keine KI-Modelle mit Ihren Daten. Wir verwenden Ihre Unterhaltungen, Ihre Dateien, Ihre Präferenzen oder Daten aus Ihren verbundenen Drittdiensten nicht, um KI-Modelle zu trainieren, zu optimieren (Fine-Tuning) oder zu verbessern. Weder unsere eigenen noch die anderer. Dies ist eine grundlegende Verpflichtung von eustella und nicht bloß eine Einstellung, die Sie erst finden und aktivieren müssen. Wir verkaufen Ihre Daten nicht. Wir verkaufen, vermieten oder handeln Ihre personenbezogenen Daten nicht — an niemanden und aus keinem Grund. Wir verwenden Ihre Daten nicht für Werbung. eustella schaltet keine Werbung, und wir geben Ihre Daten nicht an Werbetreibende oder Werbenetzwerke weiter. Wir übermitteln Ihre Daten nicht außerhalb der Europäischen Union. Unsere gesamte Infrastruktur, Datenverarbeitung und Speicherung findet innerhalb der EU statt. Ihre Daten verlassen die Europäische Union nicht. Das bedeutet, dass Ihre Daten jederzeit vollständig durch die DSGVO geschützt sind und nicht den Gesetzen eines Nicht-EU-Landes unterliegen — einschließlich des US CLOUD Act. ## Datenschutz › eustella Datenschutzerklärung eustella lernt Ihre Präferenzen, um Ihnen bessere Antworten zu geben. Dies ist eine Kernfunktion des Dienstes und funktioniert wie folgt. Woran sich eustella erinnert. Während Sie eustella nutzen, kann sie Informationen über Ihre Präferenzen speichern — zum Beispiel, dass Sie prägnante Antworten bevorzugen, dass Sie in einer bestimmten Branche tätig sind, dass Sie einen bestimmten Kommunikationsstil mögen oder dass Sie sie gebeten haben, sich eine bestimmte Tatsache über Ihr Leben zu merken. Diese Informationen werden als Teil Ihres Kontos gespeichert und ausschließlich dazu verwendet, die Antworten von eustella für Sie zu personalisieren. Sie haben stets die Kontrolle. Sie können jederzeit in Ihren Kontoeinstellungen alles einsehen, was eustella über Sie gespeichert hat. Sie können jede einzelne Präferenz bearbeiten oder löschen. Sie können außerdem die gesamte Personalisierungsfunktion deaktivieren; in diesem Fall speichert eustella keine Präferenzen, und jede Unterhaltung beginnt von Neuem. Dies ist keine automatisierte Entscheidungsfindung. Nach der DSGVO (Art. 22) haben Sie das Recht, nicht einer ausschließlich auf automatisierter Verarbeitung beruhenden Entscheidung unterworfen zu werden, die Ihnen gegenüber rechtliche Wirkung entfaltet oder Sie in ähnlicher Weise erheblich beeinträchtigt. Die Personalisierung von eustella trifft keinerlei Entscheidungen über Sie. Sie bestimmt nicht Ihre Berechtigung für irgendetwas, schränkt Ihren Zugang zu keinem Dienst ein, bewertet nicht Ihre Kreditwürdigkeit, beurteilt nicht Ihre Arbeitsleistung und berührt Ihre gesetzlichen Rechte in keiner Weise. Sie passt lediglich die Art an, wie sie mit Ihnen spricht — die Worte, die sie wählt, den Detailgrad, den sie bietet, die Themen, an die sie sich erinnert —, und zwar auf Grundlage dessen, was Sie ihr mitgeteilt haben oder was sie aus Ihren Unterhaltungen gelernt hat. Auf Grundlage dieser Informationen wird keine Entscheidung mit rechtlicher oder ähnlich erheblicher Wirkung getroffen. Bitte beachten Sie, dass die Personalisierung von eustella aufgrund ihrer Natur beeinflussen kann, welche Informationen Ihnen präsentiert werden. Dies ist kein Profiling für Zwecke Dritter. eustella erstellt keine Profile über Sie für Werbung, Marketing, Kreditbewertung, Versicherungsbeurteilung oder irgendeinen externen Zweck. Ihre Präferenzen bestehen ausschließlich zwischen Ihnen und eustella und werden nur dazu verwendet, den Dienst für Sie nützlicher zu machen. Kurz gesagt: eustella merkt sich, was Sie ihr sagen, damit sie Ihnen besser helfen kann. Sie können genau sehen, was sie weiß, es ändern oder es abschalten. Mehr tut sie nicht. ## Datenschutz › eustella Datenschutzerklärung Die DSGVO gewährt bestimmten besonders sensiblen Arten personenbezogener Daten zusätzlichen Schutz — darunter Gesundheitsdaten, Daten über religiöse oder weltanschauliche Überzeugungen, politische Meinungen, die sexuelle Orientierung, die rassische oder ethnische Herkunft sowie genetische oder biometrische Daten (Art. 9 DSGVO). eustella ist nicht darauf ausgelegt, sensible Daten zu verarbeiten. Unser Dienst ist ein KI-Assistent für allgemeine Zwecke. Wir fragen Sie nicht nach sensiblen Daten, wir setzen sie nicht voraus, und unsere Funktionen sind nicht auf sie angewiesen. Wir können Sie jedoch nicht daran hindern, sie zu teilen. Da eustella ein dialogorientiertes Werkzeug ist, können Sie alles eingeben. Wir können Sie nicht daran hindern, eustella von einem Gesundheitszustand zu erzählen, sie in einem persönlichen Kontext zu einem religiösen Thema zu fragen oder andere Informationen zu teilen, die unter eine besondere Kategorie fallen. Wir möchten hierüber ehrlich sein, statt so zu tun, als könne dies nicht geschehen. Unser dringender Rat. Bitte teilen Sie sensible personenbezogene Daten nur dann mit eustella, wenn Sie sie für den Zweck Ihrer Unterhaltung wirklich benötigen. Teilen Sie insbesondere keine sensiblen personenbezogenen Daten über andere Personen. Wenn Sie sensible Daten teilen, gehen wir wie folgt damit um. Wir verarbeiten sie nur in dem Umfang, der von der Bereitstellung der von Ihnen angeforderten Dienste untrennbar ist, und wenden die folgenden Schutzmaßnahmen an. Wir geben sensible Daten nicht zu Marketing-, Werbe- oder Profiling-Zwecken an Dritte weiter. Wir verwenden sensible Daten nicht, um Profile über Sie zu erstellen oder irgendeine Beurteilung über Sie vorzunehmen. Wir verwenden sensible Daten nicht, um KI-Modelle zu trainieren. Wenn die Personalisierung aktiviert ist, kann sich eustella an Kontext aus Ihren Unterhaltungen erinnern, der sensible Informationen enthalten könnte, die Sie angegeben haben — Sie können dies jederzeit einsehen und löschen. Wir wenden auf alle Daten, einschließlich sensibler Daten, dieselben Sicherheitsmaßnahmen an (Verschlüsselung im Ruhezustand und bei der Übertragung, Zugriffskontrollen). Sie können einzelne Unterhaltungen oder Ihr gesamtes Konto jederzeit löschen. Ein Hinweis zu Daten über andere Personen. Bitte geben Sie keine personenbezogenen Daten Dritter in eustella ein — insbesondere keine sensiblen Daten —, es sei denn, Sie haben eine Rechtsgrundlage für deren Weitergabe. Sie sind dafür verantwortlich sicherzustellen, dass Sie berechtigt sind, alle von Ihnen bereitgestellten personenbezogenen Daten über andere Personen zu teilen. ## Datenschutz › eustella Datenschutzerklärung Wir betreiben eigene Sicherheitssysteme, um Missbrauch von eustella zu erkennen und zu verhindern. Dies ist zur Erfüllung unserer rechtlichen Verpflichtungen erforderlich und zugleich etwas, das wir für richtig halten. Was das in der Praxis bedeutet. Unsere Sicherheitssysteme können den Inhalt Ihrer Unterhaltungen analysieren, um Verstöße gegen unsere Nutzungsrichtlinien zu erkennen — etwa Versuche, illegale Inhalte zu erzeugen, Inhalte, die Minderjährigen schaden könnten, oder Inhalte, die Gewalt oder andere schwerwiegende Schäden erleichtern würden. Diese Analyse erfolgt automatisiert und ist darauf ausgelegt, mögliche Verstöße zu kennzeichnen. Wird eine Kennzeichnung ausgelöst, kann eine begrenzte Anzahl geschulter Sicherheitsmitarbeiter die gekennzeichneten Inhalte überprüfen. Was das nicht bedeutet. Inhaltsmoderation ist keine allgemeine Überwachung Ihrer Unterhaltungen. Wir lesen Ihre Unterhaltungen nicht standardmäßig. Wir verwenden Daten aus der Inhaltsmoderation zu keinem anderen Zweck als Sicherheit und Rechtskonformität. Mitarbeiter, die Sicherheitsüberprüfungen durchführen, unterliegen strengen Vertraulichkeitspflichten und Zugriffskontrollen. Die Rechtsgrundlage für diese Verarbeitung ist unser berechtigtes Interesse an der Aufrechterhaltung eines sicheren und rechtmäßigen Dienstes (Art. 6 Abs. 1 lit. f DSGVO). ## Datenschutz › eustella Datenschutzerklärung Wir halten die Zahl der Parteien, die Zugang zu Ihren Daten haben, auf ein absolutes Minimum. Hier die Kategorien von Empfängern und warum sie Daten erhalten. Infrastruktur- und Hosting-Anbieter — Wir nutzen in der EU ansässige Cloud-Infrastruktur-Anbieter, um eustella zu hosten und Ihre Daten zu speichern. Diese Anbieter verarbeiten Daten ausschließlich nach unseren Weisungen und sind durch Auftragsverarbeitungsverträge gebunden. Das gesamte Hosting findet innerhalb der EU statt. KI-Modell-Anbieter — eustella verwendet quelloffene große Sprachmodelle sowie ausgewählte in der EU ansässige KI-Verarbeitungsanbieter, um Antworten zu erzeugen. Diese Modelle werden ausschließlich auf in der EU ansässiger Infrastruktur gehostet. Ihre Unterhaltungen werden über diese Systeme ausschließlich zur Erzeugung von Antworten verarbeitet. PostHog (Analyse) — Wir verwenden PostHog, gehostet in der EU (Frankfurt, Deutschland), für Produktanalysen. PostHog verarbeitet pseudonymisierte Nutzungsdaten in unserem Auftrag. PostHog ist unser Auftragsverarbeiter und durch einen Auftragsverarbeitungsvertrag gebunden. Zahlungsdienstleister — Wenn Sie ein kostenpflichtiges Abonnement erwerben, wird Ihre Zahlung von unserem Zahlungsdienstleister abgewickelt. Wir speichern Ihre vollständige Kreditkartennummer oder Zahlungsdaten nicht selbst. Der Zahlungsdienstleister erhält nur die Daten, die zur Abwicklung Ihrer Transaktion erforderlich sind. Rechts- und Aufsichtsbehörden — Wir können Daten an Strafverfolgungs- oder Aufsichtsbehörden weitergeben, wenn wir gesetzlich dazu verpflichtet sind, etwa auf einen gültigen Gerichtsbeschluss oder ein verbindliches rechtliches Ersuchen hin. Wir prüfen stets die Rechtsgrundlage eines solchen Ersuchens und benachrichtigen Sie, soweit gesetzlich zulässig. Fachliche Berater — In seltenen Fällen können wir Daten an unsere Rechts-, Steuer- oder Prüfungsberater weitergeben, soweit dies zum Schutz unserer rechtlichen Interessen oder zur Erfüllung unserer Pflichten erforderlich ist. Diese Berater unterliegen beruflichen Verschwiegenheitspflichten. Wir führen unter https://eustella.com/services/privacy/sub-processors eine aktuelle Liste unserer Unterauftragsverarbeiter. Wenn wir wesentliche Änderungen an dieser Liste vornehmen, benachrichtigen wir Sie mindestens 30 Tage im Voraus per E-Mail oder über eine In-App-Benachrichtigung. ## Datenschutz › eustella Datenschutzerklärung Wir übermitteln Ihre personenbezogenen Daten nicht außerhalb der Europäischen Union. Die gesamte Datenverarbeitung, -speicherung und das Hosting finden innerhalb der EU statt. Dies gilt für Ihre Unterhaltungen, Ihre Kontodaten, Ihre Präferenzen, Ihre Analysedaten und alle Daten aus verbundenen Drittdiensten. Das bedeutet, dass Ihre Daten ausschließlich dem EU-Datenschutzrecht unterliegen und nicht den Rechtsrahmen eines Nicht-EU-Landes ausgesetzt sind. Wenn Sie Drittdienste wie Google Calendar oder Google Drive verbinden, werden Daten aus diesen Diensten in unsere in der EU ansässige Infrastruktur abgerufen und ausschließlich innerhalb der EU verarbeitet. Ihre Daten innerhalb dieser Drittdienste unterliegen jedoch weiterhin deren eigenen Datenschutzerklärungen und Datenübermittlungspraktiken, die auch Übermittlungen außerhalb der EU umfassen können. Unsere Zusage der Nichtübermittlung gilt für unsere eigenen Systeme und unsere Infrastruktur — wir haben keinen Einfluss darauf, wie Drittdienste Daten auf ihrer Seite handhaben. ## Datenschutz › eustella Datenschutzerklärung Wir bewahren Ihre Daten nur so lange auf, wie es für den Zweck erforderlich ist, für den sie erhoben wurden. Hier die konkreten Aufbewahrungsfristen. Unterhaltungsdaten — Ihre Unterhaltungen werden gespeichert, solange Ihr Konto besteht. Sie können einzelne Unterhaltungen jederzeit löschen; gelöschte Unterhaltungen werden innerhalb von 30 Tagen aus unseren aktiven Systemen entfernt. Wenn Sie Ihr Konto löschen, werden alle Unterhaltungsdaten innerhalb von 30 Tagen gelöscht. Personalisierungspräferenzen — Ihre gespeicherten Präferenzen werden aufbewahrt, solange Ihr Konto besteht und die Personalisierungsfunktion aktiviert ist. Sie können einzelne Präferenzen jederzeit löschen oder die Personalisierung vollständig deaktivieren. Wenn Sie Ihr Konto löschen, werden alle Präferenzen innerhalb von 30 Tagen gelöscht. Kontodaten (Name, E-Mail, Kontoeinstellungen) — Aufbewahrt für die Dauer Ihres Kontos zuzüglich 3 Jahre nach Kontoschließung, um die gesetzlichen Verjährungsfristen nach österreichischem Recht (ABGB) abzudecken. Rechnungs- und Zahlungsdaten — Aufbewahrt für 7 Jahre nach der betreffenden Transaktion, wie nach österreichischem Rechnungslegungsrecht (Unternehmensgesetzbuch / Bundesabgabenordnung) vorgeschrieben. Sicherheits- und Protokolldaten — Aufbewahrt für 30 bis 90 Tage, je nach Art des Protokolls. Diese Daten werden ausschließlich für Sicherheitsüberwachung und Vorfallsreaktion verwendet. Aufzeichnungen zur Inhaltsmoderation — Wenn Inhalte von unseren Sicherheitssystemen gekennzeichnet wurden, wird ein Datensatz über die Kennzeichnung und etwaige ergriffene Maßnahmen für bis zu 12 Monate aufbewahrt, zum Zweck der Durchsetzung unserer Richtlinien und der Einhaltung einschlägiger Gesetze und Vorschriften. Backups — Unsere Backup-Systeme bewahren Daten für bis zu 90 Tage auf. Wenn Daten aus unseren aktiven Systemen gelöscht werden, werden sie innerhalb dieses Backup-Rotationszeitraums aus den Backups entfernt. Analysedaten — Über PostHog erhobene Nutzungsanalysen werden 3 Jahre lang aufbewahrt. Diese Daten sind pseudonymisiert und enthalten nicht den Inhalt Ihrer Unterhaltungen. ## Datenschutz › eustella Datenschutzerklärung Nach der DSGVO haben Sie die folgenden Rechte. Sie können sie jederzeit ausüben, indem Sie sich an privacy@eustella.com wenden oder die entsprechenden Funktionen in Ihren Kontoeinstellungen nutzen. Wir antworten innerhalb eines Monats, wie gesetzlich vorgeschrieben. Recht auf Auskunft — Sie können eine Kopie der personenbezogenen Daten anfordern, die wir über Sie führen. Wir stellen sie in einem gängigen, maschinenlesbaren Format bereit. Recht auf Berichtigung — Wenn Ihre Daten unrichtig sind, können Sie uns bitten, sie zu berichtigen. Bitte beachten Sie, dass eustella ein KI-System ist und mitunter unrichtige Informationen erzeugen kann (sogenannte „Halluzinationen"). Wenn eustella eine falsche Präferenz über Sie gespeichert hat, können Sie diese direkt in Ihren Kontoeinstellungen korrigieren. Wenn eustella innerhalb einer Unterhaltung unrichtige Inhalte über Sie erzeugt hat, können Sie diese Unterhaltung löschen. Recht auf Löschung — Sie können uns bitten, Ihre personenbezogenen Daten zu löschen. Sie können außerdem einzelne Unterhaltungen, einzelne Präferenzen oder Ihr gesamtes Konto jederzeit über die App löschen. Bitte beachten Sie, dass wir bestimmte Daten möglicherweise aufbewahren müssen, sofern dies gesetzlich vorgeschrieben ist (zum Beispiel Rechnungsunterlagen nach österreichischem Rechnungslegungsrecht). Recht auf Einschränkung der Verarbeitung — Sie können uns unter bestimmten Umständen bitten, die Verwendung Ihrer Daten einzuschränken, etwa während wir die Richtigkeit von Daten überprüfen, die Sie bestritten haben. Recht auf Datenübertragbarkeit — Sie können verlangen, dass wir Ihre Daten in einem strukturierten, gängigen, maschinenlesbaren Format bereitstellen oder sie, soweit technisch machbar, direkt an einen anderen Verantwortlichen übermitteln. Widerspruchsrecht — Sie haben das Recht, einer Verarbeitung zu widersprechen, die auf unserem berechtigten Interesse beruht (Art. 6 Abs. 1 lit. f DSGVO), einschließlich wesentlicher Analysen. Sie können widersprechen, ohne einen Grund angeben zu müssen, und wir stellen die Verarbeitung Ihrer Daten ein, es sei denn, wir können zwingende schutzwürdige Gründe nachweisen, die Ihre Rechte überwiegen. Recht auf Widerruf der Einwilligung — Soweit wir uns auf Ihre Einwilligung als Rechtsgrundlage stützen (zum Beispiel für Marketingmitteilungen und nicht wesentliche Analysen), können Sie diese Einwilligung jederzeit widerrufen. Der Widerruf der Einwilligung berührt nicht die Rechtmäßigkeit der vor dem Widerruf erfolgten Verarbeitung. Recht auf Beschwerde — Wenn Sie mit unserem Umgang mit Ihren Daten nicht zufrieden sind, haben Sie das Recht, Beschwerde bei einer Aufsichtsbehörde einzulegen. In Österreich ist die zuständige Behörde: Österreichische Datenschutzbehörde Barichgasse 40–42, 1030 Wien, Österreich E-Mail: dsb@dsb.gv.at Website: https://www.dsb.gv.at Wenn Sie in einem anderen EU-Mitgliedstaat ansässig sind, können Sie auch Beschwerde bei der Aufsichtsbehörde Ihres Wohnsitzlandes einlegen. Ist die Bereitstellung Ihrer Daten erforderlich? Einige Daten sind für die Nutzung von eustella erforderlich. Sie können kein Konto erstellen, ohne Ihren Namen und Ihre E-Mail-Adresse anzugeben. Sie können den Dienst nicht nutzen, ohne dass Ihre Eingaben zur Erzeugung von Antworten verarbeitet werden — so funktioniert eine dialogorientierte KI. Wenn Sie sich entscheiden, diese Daten nicht bereitzustellen, können Sie eustella nicht nutzen. Andere Daten sind optional. Sie sind nicht verpflichtet, Drittdienste zu verbinden, die Personalisierung zu aktivieren oder eine bestimmte Art von Informationen in Ihren Unterhaltungen zu teilen. Wenn Sie dies nicht tun, können Sie eustella dennoch nutzen — Sie haben lediglich ein weniger personalisiertes Erlebnis. Recht in Bezug auf automatisierte Entscheidungsfindung — Sie haben das Recht, nicht einer ausschließlich auf automatisierter Verarbeitung — einschließlich Profiling — beruhenden Entscheidung unterworfen zu werden, die Ihnen gegenüber rechtliche Wirkung entfaltet oder Sie in ähnlicher Weise erheblich beeinträchtigt (Art. 22 DSGVO). eustella trifft keine solchen Entscheidungen. Ihre Präferenzen werden ausschließlich zur Personalisierung Ihres Erlebnisses gespeichert, und Sie haben jederzeit die volle Kontrolle darüber, wie in Abschnitt 5 beschrieben. ## Datenschutz › eustella Datenschutzerklärung eustella ist nicht für die Nutzung durch Personen unter 16 Jahren bestimmt. Wir haben 16 als unser Mindestalter gewählt, weil dies das Standardalter für die digitale Einwilligung nach der DSGVO ist, wodurch die Konformität in allen EU-Mitgliedstaaten gewährleistet wird. Wir erkennen an, dass mit den derzeit verfügbaren technischen Mitteln das Alter einer Person nicht vollständig überprüft werden kann, ohne unverhältnismäßig in ihre Privatsphäre einzugreifen. Wir beobachten Entwicklungen in diesem Bereich aktiv und werden verhältnismäßige Überprüfungsmaßnahmen umsetzen, sobald sie verfügbar werden und sich der regulatorische Rahmen weiterentwickelt. Was geschieht, wenn ein Minderjähriger eustella nutzt. Wenn wir darüber informiert werden, dass eine Person unter 16 Jahren eustella nutzt, werden wir das Konto (sofern vorhanden) umgehend sperren und alle mit dieser Person verbundenen personenbezogenen Daten löschen. Die Löschung aus unseren aktiven Systemen erfolgt innerhalb von 30 Tagen nach Erhalt der Mitteilung. Die Löschung aus Backup-Systemen erfolgt bei der nächsten planmäßigen Backup-Rotation, spätestens innerhalb von 90 Tagen. Wie Sie dies melden. Wenn Sie glauben, dass eine Person unter 16 Jahren eustella nutzt, kontaktieren Sie uns bitte unter privacy@eustella.com. Wir bemühen uns, jede Meldung innerhalb von 72 Stunden zu bestätigen und wie oben beschrieben zu handeln. Für Nutzer in den Vereinigten Staaten. Sollten wir unseren Dienst künftig auf die Vereinigten Staaten ausweiten, werden wir die geltenden US-Gesetze einhalten, einschließlich des COPPA (Children's Online Privacy Protection Act). ## Datenschutz › eustella Datenschutzerklärung Wir nehmen die Sicherheit Ihrer Daten ernst und wenden angemessene technische und organisatorische Maßnahmen an, darunter die Verschlüsselung von Daten im Ruhezustand und bei der Übertragung (unter Verwendung branchenüblicher Protokolle wie AES-256 und TLS 1.2 oder höher), strenge Zugriffskontrollen (kein Mitarbeiter hat standardmäßigen Zugriff auf den Inhalt Ihrer Unterhaltungen — der Zugriff ist auf benannte Sicherheitsmitarbeiter unter bestimmten, dokumentierten Umständen beschränkt), regelmäßige Sicherheitsbewertungen und Pseudonymisierung, soweit technisch machbar (zum Beispiel die Trennung von Unterhaltungsinhalten und Kontokennungen in unseren internen Systemen). Kein System kann absolute Sicherheit garantieren. Wenn Sie auf ein Sicherheitsproblem im Zusammenhang mit eustella aufmerksam werden, kontaktieren Sie uns bitte umgehend unter security@eustella.com. ## Datenschutz › eustella Datenschutzerklärung Wir können diese Datenschutzerklärung von Zeit zu Zeit aktualisieren, wenn wir neue Funktionen entwickeln oder sich die Rechtslage ändert. Wenn wir Änderungen vornehmen, die Ihre Rechte oder die Art, wie wir Ihre Daten verwenden, wesentlich betreffen, benachrichtigen wir Sie mindestens 30 Tage im Voraus, entweder per E-Mail oder über eine In-App-Benachrichtigung. Wir werden Ihre Rechte nach dieser Datenschutzerklärung nicht ohne Ihre ausdrückliche Einwilligung einschränken. Die aktuelle Fassung dieser Datenschutzerklärung ist stets hier verfügbar. ## Datenschutz › eustella Datenschutzerklärung Für alle Fragen, Anliegen oder Anträge im Zusammenhang mit Ihrer Privatsphäre oder dieser Datenschutzerklärung: Datenschutz-Kontakt: privacy@eustella.com Sicherheitsbedenken: security@eustella.com Allgemeine rechtliche Anfragen: legal@eustella.com Postanschrift: AI Newsrooms Technology GmbH Karl-Farkas-Gasse 22, 1030 Wien, Österreich © 2026 AI Newsrooms Technology GmbH. Alle Rechte vorbehalten. --- # eustella — Privacy URL: https://eustella.com/services/privacy Topic: service-privacy ## Privacy Last updated: 22.06.2026 This Privacy Policy explains how we collect, use, and protect your personal data when you use eustella. We wrote it to be clear and honest - if not, please contact us and we will be happy to explain. This Privacy Policy forms part of your relationship with us but is separate from our Terms of Use. It applies to your use of eustella's mobile applications, web interfaces, AI agent features, and all related services ("Services"). If you are looking for the Privacy Policy regarding our website eustella.com, please visit eustella.com/privacy. ## Privacy › eustella Privacy Policy eustella is built by AI Newsrooms Technology GmbH, a European AI company based in Vienna, Austria. We are the controller of your personal data, which means we decide how and why your data is processed. AI Newsrooms Technology GmbH Karl-Farkas-Gasse 22, 1030 Vienna, Austria Company register: 637289s, Commercial Court of Vienna Email: privacy@eustella.com If you have any questions about how we handle your data, please contact us at privacy@eustella.com. We aim to respond within 14 days. ## Privacy › eustella Privacy Policy We collect different types of data depending on how you use eustella. Here is an overview. Data you give us directly When you create an account, we collect your name, email address, and password. If you sign up through Google or Apple, we receive the profile information those services share with us (typically your name and email address). We do not receive or store your Google or Apple password. When you use eustella, we collect the content of your conversations - your inputs (what you type or upload) and eustella's outputs (what eustella generates for you). If you upload files, images, or documents, we process those too. If you connect third-party services such as Google Calendar or Google Drive, we access and process data from those services on your behalf and only to the extent necessary to provide the features you have enabled. We do not access data from your connected services for any other purpose. Data we collect automatically When you use our Services, we automatically collect certain technical and usage data. This includes your IP address, device type, operating system, browser type, language settings, and general location (derived from your IP address at the country or city level - we do not collect precise GPS location data). We also collect information about how you use eustella, such as which features you use, when you use them, and how you interact with the app. This data is collected through PostHog, our analytics provider, which is hosted in the European Union (Frankfurt, Germany). Data eustella learns about you eustella is a personalised AI assistant. One of its core features is learning your preferences over time so it can give you more useful and relevant responses. This means eustella may remember things like your preferred language, your communication style, the topics you are interested in, or the way you like information presented. We explain exactly how this works and what control you have in Section 5. ## Privacy › eustella Privacy Policy The GDPR requires us to have a specific legal reason for every way we use your personal data. Here is an overview. Providing and operating eustella - When we process your conversations, generate responses, connect to your third-party services, save your preferences (as described in Section 5), and generally make eustella work for you, we do so because it is necessary to perform our contract with you (Art. 6(1)(b) GDPR). You asked us to provide these services, and we need to process your data to do so. Keeping eustella safe - We operate our own content moderation and safety systems to detect misuse, enforce our Usage Policies, and protect our users and the public. We do this based on our legitimate interest in maintaining a safe and lawful service (Art. 6(1)(f) GDPR), and where required, to comply with our legal obligation under applicbale laws and regualtions, including the Digital Services Act (Art. 6(1)(c) GDPR). Improving our Services - We use aggregated and anonymised usage data (such as which features are popular, where users encounter errors, and general usage patterns) to improve eustella. We rely on our legitimate interest in developing and improving our service (Art. 6(1)(f) GDPR). We do not use the content of your conversations to train or improve AI models. Analytics - We use PostHog (hosted in the EU) to understand how people use eustella so we can identify issues and improve the experience. For essential analytics we rely on our legitimate interest in reliably operating and improving our service (Art. 6(1)(f) GDPR). For non-essential analytics we rely on your consent (Art. 6(1)(a) GDPR), which you can give or withdraw at any time in your account settings or by contacting us. Withdrawing consent does not affect the lawfulness of processing that took place before withdrawl. Communicating with you - We may send you service-related communications, such as security alerts, changes to our terms, or important updates about your account. These are necessary for the performance of our contract with you (Art. 6(1)(b) GDPR). If we ever send you marketing communications, we will only do so with your explicit consent (Art. 6(1)(a) GDPR), and you can withdraw that consent at any time. Complying with the law - We process data when required by law, for example to respond to valid legal requests from authorities or to fulfil our tax and accounting obligations under Austrian law (Art. 6(1)(c) GDPR). Protecting our rights - In rare cases, we may process data to establish, exercise, or defend legal claims. We rely on our legitimate interest in protecting our legal position (Art. 6(1)(f) GDPR). ## Privacy › eustella Privacy Policy Some things are just as important to say plainly. We do not train AI models on your data. We do not use your conversations, your files, your preferences, or any data from your connected third-party services to train, fine-tune, or improve AI models. Not ours, not anyone else's. This is a core commitment of eustella, not just a setting you need to find and toggle. We do not sell your data. We do not sell, rent, or trade your personal data to anyone, for any reason. We do not use your data for advertising. eustella does not serve advertisements and we do not share your data with advertisers or ad networks. We do not transfer your data outside the European Union. All of our infrastructure, data processing, and storage takes place within the EU. Your data does not leave the European Union. This means your data is is fully protected by the GDPR at all times and not subject to the laws of any non-EU country - including the U.S. CLOUD Act. ## Privacy › eustella Privacy Policy eustella learns your preferences to give you better responses. This is a core feature of the service and works as follows. What eustella remembers. As you use eustella, it may save information about your preferences - for example, that you prefer concise answers, that you work in a particular industry, that you like a certain communication style, or that you have asked it to remember a specific fact about your life. This information is stored as part of your account and is used exclusively to personalise eustella's responses to you. You are always in control. You can view everything eustella has saved about you at any time in your account settings. You can edit or delete any individual preference. You can also turn off the entire personalisation feature, in which case eustella will not save any preferences and each conversation will start fresh. This is not automated decision-making. Under the GDPR (Art. 22), you have the right not to be subject to decisions based solely on automated processing that produce legal effects or similarly significant effects on you. eustella's personalisation does not make any decisions about you. It does not determine your eligibility for anything, restrict your access to any service, assess your creditworthiness, evaluate your job performance, or affect your legal rights in any way. It adapts the way it talks to you - the words it chooses, the level of detail it provides, the topics it remembers - based on what you have told it or what it has learned from your conversations. No decision with legal or similarly significant effect is made on the basis of this information. Keep in mind that the personalization of eustella may effect what information is presented to you due to the nature of personalization. This is not profiling for third-party purposes. eustella does not build profiles about you for advertising, marketing, credit scoring, insurance assessment, or any external purpose. Your preferences exist solely between you and eustella and are used only to make the service more useful to you. In short: eustella remembers what you tell it so it can help you better. You can see exactly what it knows, change it, or turn it off. That is all it does. ## Privacy › eustella Privacy Policy The GDPR gives additional protection to certain especially sensitive types of personal data - including health information, data about religious or philosophical beliefs, political opinions, sexual orientation, racial or ethnic origin, and genetic or biometric data (Art. 9 GDPR). eustella is not designed to process sensitive data. Our service is a general-purpose AI assistant. We do not ask you for sensitive data, we do not require it, and our features do not depend on it. But we cannot prevent you from sharing it. Because eustella is a conversational tool, you can type anything into it. We cannot prevent you from telling eustella about a health condition, ask it about a religious topic in a personal context, or share other information that falls into a special category. We want to be honest about this reality rather than pretend it cannot happen. Our strong advise. Please do not share sensitive personal data with eustella unless you genuinely need to for the purpose of your conversation. In particular, please do not share sensitive personal data about other people. If you do share sensitive data, here is how we handle it. We process it only to the extent inseparable from providing the services you requested, and we apply the following safeguards. We do not share sensitive data with any third party for marketing, advertising, or profiling purposes. We do not use sensitive data to build profiles about you or to make any kind of assessment about you. We do not use sensitive data to train AI models. If personalisation is enabled, eustella may remember context from your conversations, which could include sensitive information you provided - you can review and delete this at any time. We apply the same security measures (encryption at rest and in transit, access controls) to all data, including sensitive data. You can delete individual conversations or your entire account at any time. A note on data about other people. Please do not enter the personal data of third parties into eustella - particularly not sensitive data - unless you have a lawful basis for sharing it. You are responsible for ensuring that you have the right to share any personal data you provide about other people. ## Privacy › eustella Privacy Policy We operate our own safety systems to detect and prevent misuse of eustella. This is required to comply with our legal obligations and is also something we believe is the right thing to do. What this means in practice. Our safety systems may analyse the content of your conversations to identify violations of our Usage Policies - for example, attempts to generate illegal content, content that could harm minors, or content that would facilitate violence or other serious harms. This analysis is automated and is designed to flag potential violations. Where a flag is raised, a limited number of trained safety staff may review the flagged content. What this does not mean. Content moderation is not a general surveillance of your conversations. We do not read your conversations by default. We do not use content moderation data for any purpose other than safety and compliance. Staff who conduct safety reviews are bound by strict confidentiality obligations and access controls. The legal basis for this processing is our legitimate interest in maintaining a safe and lawful service (Art. 6(1)(f) GDPR). ## Privacy › eustella Privacy Policy We keep the number of parties who have access to your data to an absolute minimum. Here are the categories of recipients and why they receive data. Infrastructure and hosting providers - We use EU-based cloud infrastructure providers to host eustella and store your data. These providers process data solely on our instructions and are bound by data processing agreements. All hosting takes place within the EU. AI model providers - eustella uses open-source large language models and selected EU-based AI processing providers to generate responses. These models are solely hosted on EU-based infrastructure. Your conversations are processed through these systems only to generate responses. PostHog (analytics) - We use PostHog, hosted in the EU (Frankfurt, Germany), for product analytics. PostHog processes pseudonymised usage data on our behalf. PostHog is our data processor and is bound by a data processing agreement. Payment providers - If you purchase a paid subscription, your payment is processed by our payment provider. We do not store your full credit card number or payment details ourselves. The payment provider receives only the data necessary to process your transaction. Legal and regulatory authorities - We may share data with law enforcement or regulatory authorities when we are legally required to do so, for example in response to a valid court order or binding legal request. We will always verify the legal basis of any such request and, where permitted by law, we will notify you. Professional advisors - In rare cases, we may share data with our legal, tax, or audit advisors where necessary to protect our legal interests or comply with our obligations. These advisors are bound by professional confidentiality obligations. We maintain a current list of our sub-processors at https://eustella.com/services/privacy/sub-processors. If we make significant changes to this list, we will notify you at least 30 days in advance by email or through an in-app notification. ## Privacy › eustella Privacy Policy We do not transfer your personal data outside the European Union. All data processing, storage, and hosting takes place within the EU. This applies to your conversations, your account data, your preferences, your analytics data, and any data from connected third-party services. This means your data is subject exclusively to EU data protection law and is not exposed to the legal frameworks of any non-EU country. When you connect third-party services such as Google Calendar or Google Drive, data from those services is retrieved into our EU-based infrastructure and processed exclusively within the EU. However, your data within those third-party services remains subject to their own privacy policies and data transfer practices, which may include transfers outside the EU. Our no-transfer commitment applies to our own systems and infrastructure - we cannot control how third-party services handle data on their end. ## Privacy › eustella Privacy Policy We retain your data only for as long as necessary for the purpose for which it was collected. Here are the specific retention periods. Conversation data - Your conversations are stored for as long as your account exists. You can delete individual conversations at any time, and deleted conversations are removed from our active systems within 30 days. If you delete your account, all conversation data is deleted within 30 days. Personalisation preferences - Your saved preferences are stored for as long as your account exists and the personalisation feature is enabled. You can delete individual preferences at any time or turn off personalisation entirely. If you delete your account, all preferences are deleted within 30 days. Account data (name, email, account settings) - Retained for the duration of your account plus 3 years after account closure, to cover statutory limitation periods under Austrian law (ABGB). Billing and payment data - Retained for 7 years after the relevant transaction, as required by Austrian accounting law (Unternehmensgesetzbuch / Bundesabgabenordnung). Security and log data - Retained for 30 to 90 days, depending on the type of log. This data is used exclusively for security monitoring and incident response. Content moderation records - Where content has been flagged by our safety systems, a record of the flag and any action taken is retained for up to 12 months for the purpose of enforcing our policies and complying with relevant laws and regulations. Backups - Our backup systems retain data for up to 90 days. When data is deleted from our active systems, it will be removed from backups within this backup rotation period. Analytics data - Usage analytics collected through PostHog are retained for 3 years. This data is pseudonymised and does not include the content of your conversations. ## Privacy › eustella Privacy Policy Under the GDPR, you have the following rights. You can exercise them at any time by contacting privacy@eustella.com or using the controls in your account settings. We will respond within one month, as required by law. Right to access - You can request a copy of the personal data we hold about you. We will provide it in a commonly used, machine-readable format. Right to rectification - If your data is inaccurate, you can ask us to correct it. Please note that eustella is an AI system and may sometimes generate inaccurate information (so-called "hallucinations"). If eustella has saved an incorrect preference about you, you can correct it directly in your account settings. If eustella has generated inaccurate content about you within a conversation, you can delete that conversation. Right to erasure - You can ask us to delete your personal data. You can also delete individual conversations, individual preferences, or your entire account through the app at any time. Please be aware that we may need to retain certain data where required by law (for example, billing records under Austrian accounting law). Right to restrict processing - You can ask us to restrict how we use your data in certain circumstances, for example while we verify the accuracy of data you have disputed. Right to data portability - You can request that we provide your data in a structured, commonly used, machine-readable format, or that we transmit it directly to another controller where technically feasible. Right to object - You have the right to object to processing that is based on our legitimate interest (Art. 6(1)(f) GDPR), including essential analytics. You can object with no need to provide a reason and we will stop processing your data unless we can demonstrate compelling legitimate grounds that override your rights. Right to withdraw consent - Where we rely on your consent as a legal basis (for example, for marketing communications and non-essential analytics), you can withdraw that consent at any time. Withdrawing consent does not affect the lawfulness of processing that occurred before the withdrawal. Right to lodge a complaint - If you are not satisfied with how we handle your data, you have the right to lodge a complaint with a supervisory authority. In Austria, the competent authority is: Österreichische Datenschutzbehörde Barichgasse 40–42, 1030 Vienna, Austria Email: dsb@dsb.gv.at Website: https://www.dsb.gv.at If you reside in another EU member state, you can also lodge a complaint with the supervisory authority in your country of residence. Is providing your data required? Some data is necessary for you to use eustella. You cannot create an account without providing your name and email address. You cannot use the service without your inputs being processed to generate responses - that is how a conversational AI works. If you choose not to provide this data, you will not be able to use eustella. Other data is optional. You are not required to connect third-party services, enable personalisation, or share any particular type of information in your conversations. If you choose not to, you can still use eustella - you will simply have a less personalised experience. Right regarding automated decision-making - You have the right not to be subject to decisions based solely on automated processing, including profiling, that produce legal effects or similarly significantly affect you (Art. 22 GDPR). eustella does not make such decisions. Your preferences are stored solely to personalise your experience, and you have full control over them at all times, as described in Section 5. ## Privacy › eustella Privacy Policy eustella is not intended for use by anyone under the age of 16. We chose 16 as our minimum age because that is the default age of digital consent under the GDPR, ensuring compliance across all EU member states. We acknowledge that no currently available technical means can fully verify a person's age without disproportionately intruding on their privacy. We actively monitor developments in this area and will implement proportionate verification measures as they become available and as the regulatory framework evolves. What happens if a minor is using eustella. If we are notified that a person under 16 is using eustella, we will promptly suspend the account (if one exists) and delete all personal data associated with that person. Deletion from our active systems will take place within 30 days of receiving the notification. Deletion from backup systems will take place within the next scheduled backup rotation, at most within 90 days. How to report. If you believe a person under 16 is using eustella, please contact us at privacy@eustella.com. We aim to acknowledge every report within 72 hours and take action as described above. For users in the United States. If we expand our service to the United States in the future, we will comply with applicable U.S. laws including COPPA (Children's Online Privacy Protection Act). ## Privacy › eustella Privacy Policy We take the security of your data seriously and apply appropriate technical and organisational measures, including encryption of data at rest and in transit (using industry-standard protocols such as AES-256 and TLS 1.2 or higher), strict access controls (no employee has default access to the content of your conversations - access is limited to designated safety staff in specific, documented circumstances), regular security assessments, and pseudonymisation where technically feasible (for example, separating conversation content from account identifiers in our internal systems). No system can guarantee absolute security. If you become aware of any security concern related to eustella, please contact us immediately at security@eustella.com. ## Privacy › eustella Privacy Policy We may update this Privacy Policy from time to time as we develop new features or as the law changes. If we make changes that materially affect your rights or the way we use your data, we will give you at least 30 days advance notice, either by email or through an in-app notification. We will not reduce your rights under this Privacy Policy without your explicit consent. The current version of this Privacy Policy is always available here. ## Privacy › eustella Privacy Policy For any questions, concerns, or requests related to your privacy or this Privacy Policy: Privacy contact: privacy@eustella.com Security concerns: security@eustella.com General legal inquiries: legal@eustella.com Postal address: AI Newsrooms Technology GmbH Karl-Farkas-Gasse 22, 1030 Vienna, Austria © 2026 AI Newsrooms Technology GmbH. All rights reserved. --- # eustella — Sub-Processors URL: https://eustella.com/services/privacy/sub-processors Topic: subprocessors ## Sub-Processors Last updated: 22.06.2026 When you use eustella, some of your data is processed not just by us, but by a small number of carefully selected third-party companies that help us run the app. Under EU data protection law (GDPR Art. 28), we are required to be transparent about who these companies are, what they do with your data, and why we chose them. We call these companies subprocessors - they act on our instructions and are not allowed to use your data for their own purposes. We believe you deserve a plain-language explanation, not just a legal table. So below, for each company, we tell you what they do, why we need them, and - where the company is based outside the EU - why there is genuinely no European alternative that can do the same job. We update this list if we are adding or materially changing any subprocessor. If you want to be notified of changes, contact us at privacy@eustella.com This list does not apply to dedicated B2B customers with whom AI Newsrooms Technology GmbH has concluded a manual deal or partnership. For those customers, the applicable sub-processors are governed by the terms of their individually concluded agreement and the dedicated Data Processing Agreement. ## Sub-Processors › eustella Subprocessors We are an Austrian company and we take EU data sovereignty seriously. Wherever a European alternative exists and is technically and operationally viable, we use it. Our core infrastructure - the servers that run eustella, store your data, and power the AI - is entirely hosted with European companies in Europe. For a small number of services, no European company offers what we need. In those cases, we try to be as transparent as possible about why, and what legal safeguards are in place to protect your data. We actively monitor the market and commit to switching to a European provider the moment a credible one becomes available. If you know of an EU-headquartered company that could replace any of the US-based providers listed below, please tell us. We genuinely want to hear it - email us at privacy@eustella.com ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entity: IONOS SE, Germany | Data location: Germany (EU) IONOS provides the servers and managed infrastructure that eustella runs on. All of your data that eustella stores lives on IONOS infrastructure in Germany. We chose IONOS because they are a German company, fully subject to EU law, with no US parent or ownership structure that could trigger foreign government data access obligations. This was a deliberate choice over larger US-based cloud providers like AWS, Google Cloud, or Microsoft Azure - all of which, despite offering EU-based data centres, remain subject to US jurisdiction and laws like the US CLOUD Act. ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entity: Scaleway SAS, France | Data location: France (EU) Scaleway provides supplementary cloud compute capacity for Eustella's backend services, as well as infrastructure for certain AI workloads. Scaleway is a French company, part of the Iliad Group, and is fully EU-headquartered with no exposure to US jurisdiction. We use Scaleway alongside IONOS and Verda to keep both our general infrastructure and our AI processing within European hands, and to avoid over-dependence on any single vendor. All three are European by design. ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entity: Mistral AI | Data location: France (EU) Mistral AI provides API services for some file-processing (OCR) and audio-transcription workflows. When eustella needs to extract text from a file or transcribe audio, the relevant content may be sent to Mistral AI for that specific processing task. Mistral AI is a French AI company headquartered in Paris and subject to EU law. We use Mistral AI where a European provider is operationally suitable for specialised document and audio processing, keeping this processing within EU jurisdiction. ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entity: Verda Cloud Oy (formerly DataCrunch), Finland | Data location: Finland (EU) Verda provides the GPU computing power needed to run AI model inference - the part of eustella that processes and responds to your inputs using AI. Verda is a Finnish company and one of the very few European providers offering high-performance GPU infrastructure at the scale AI applications require. The AI compute market is almost entirely dominated by US providers (AWS, Google Cloud, Azure, CoreWeave). Verda was a deliberate choice to keep AI inference within EU jurisdiction. They hold ISO certification confirming their security and operational standards. ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entity: Google LLC, USA | Data location: USA / EU (EU data residency where configured)Transfer safeguard: Standard Contractual Clauses (EU SCCs 2021); Google LLC certified under the EU-US Data Privacy Framework Firebase Cloud Messaging is the service that delivers notifications to your Android device - for example, when your AI companion sends you a message. There is no EU alternative for this service, and this is not a business preference but a hard technical constraint. Google Play's rules require that all Android apps use FCM for push notifications. No third-party provider, European or otherwise, can deliver notifications directly to an Android device without going through FCM. Every Android app in the world is in the same position. The data involved is minimal: only your device's registration token (a technical identifier, not your name or email address) and the content of the notification we send you. ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entity: Google LLC, USA | Data location: USA / EU (EU data residency where configured)Transfer safeguard: Standard Contractual Clauses (EU SCCs 2021); Google LLC certified under the EU-US Data Privacy Framework If you choose to sign in to eustella using your Google account, Firebase Authentication handles that login process. Google Login is Google's own identity service - authenticating a Google account requires going through Google's infrastructure, and no EU company can provide this. There is simply no such thing as a European version of Google Login. That said, this is entirely optional: you can always register with an email address and password instead, which avoids Google's involvement altogether. The data involved when you use Google Login is your Google account ID, email address, display name, and profile picture URL. ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entity: Apple Inc., USA | Data location: USATransfer safeguard: Standard Contractual Clauses (EU SCCs 2021); Apple Inc. certified under the EU-US Data Privacy Framework APNs is the service that delivers notifications to your iPhone or iPad from eustella. The situation here is identical to Android notifications above: Apple's App Store rules require all iOS apps to use APNs for push notifications. Apple controls the entire notification pipeline at the operating system level, and no EU provider can deliver a notification to an iPhone without Apple's involvement. This constraint applies to every app on your iPhone, without exception. Apple acts purely as a delivery service. The data involved is only your device's push token (a technical identifier) and the notification content we send you. ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entity: RevenueCat, Inc., USA | Data location: USATransfer safeguard: Standard Contractual Clauses (EU SCCs 2021); RevenueCat Inc. certified under the EU-US Data Privacy Framework RevenueCat manages eustella's in-app subscription system - tracking whether you have an active subscription, validating your purchase with Apple or Google, and making sure the right features are unlocked for your account. We looked carefully for a European alternative and could not find one. The market for mobile in-app subscription management is, as of April 2026, entirely dominated by US-incorporated companies. The two most credible alternatives we evaluated - Adapty (incorporated in Delaware, USA) and Qonversion (also US-incorporated) - are both subject to US jurisdiction. No EU-headquartered company currently offers the same combination of cross-platform receipt validation, entitlement management, and subscription analytics with a production-ready mobile SDK. The data we send to RevenueCat is deliberately minimal: a pseudonymous user identifier (not your name or email), your device type, subscription status, and purchase receipt data from the App Store or Google Play. We do not send RevenueCat your name, email address, or payment card details. ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entities: Apple Inc., USA and Google LLC, USA | Data location: USATransfer safeguard: Standard Contractual Clauses (EU SCCs 2021); both certified under the EU-US Data Privacy Framework When you purchase a subscription or any paid feature inside Eustella, the actual payment is processed by Apple (on iOS) or Google (on Android) - not by us directly. There is no alternative for this. Apple and Google each operate a closed payment system on their respective platforms, and their App Store rules require that all in-app purchases go through their own payment infrastructure. This is not a choice we made - it is a platform requirement that applies to every app in the world. The only notable exception is Spotify, which after years of legal battles secured special regulatory treatment in the EU under the Digital Markets Act; standard apps, including Eustella, do not have this option. This means that when you subscribe, Apple or Google receive your payment details, billing address, and transaction records directly. We never see your payment card details - they go straight to Apple or Google, who act as the merchant of record for the transaction and handle all payment data under their own privacy policies. What we receive back from Apple or Google (via RevenueCat) is only a receipt confirming that a valid purchase was made. ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entity: Linkup Technologies SAS, France | Data location: France / EUTransfer safeguard: None required - EU-headquartered company, data processed and stored within the EU Linkup provides the real-time web search capability that allows eustella to access up-to-date information from the internet when answering your questions. We are particularly pleased to use Linkup because it is a French startup, fully incorporated and headquartered in France, with all data processed and stored within the EU. This means there are no cross-border data transfer concerns whatsoever. In a market where web search APIs are dominated by US companies - Google, Microsoft Bing, and others - Linkup is a rare example of a capable, production-ready EU alternative, and we chose it deliberately for that reason. The data Linkup receives is limited to the search queries generated by the AI on your behalf. These queries are derived from your conversation but are processed without any persistent identifier linking them back to your account. ## Sub-Processors › eustella Subprocessors › Our subprocessors Legal entity: PostHog Inc., USA | Data location: Frankfurt, Germany (EU)Transfer safeguard: None required for data transfers - all data stays within the EEAResidual risk: PostHog Inc. is a US-incorporated entity and therefore theoretically subject to US CLOUD Act requests, even for data stored in Frankfurt. This risk is documented and accepted - see explanation below. PostHog helps us understand how the app is being used - which features people find useful, where things go wrong, and how to improve the experience. PostHog is a US-headquartered company, but we specifically use their EU Cloud product, which means your data is processed and stored entirely on servers in Frankfurt, Germany and never leaves the EU. However, we want to be transparent: because PostHog Inc. is incorporated in the United States, it remains theoretically subject to US government data requests under the CLOUD Act - even for data held on EU soil. This is a known and documented residual risk. We looked for a fully EU-headquartered alternative. The honest answer is that the mobile app analytics market has a significant gap: no EU-headquartered company currently offers the combination of mobile SDK quality, session analysis, feature flags, and event-level product analytics that a modern app requires. The closest EU options - such as self-hosted Matomo (open source, full sovereignty) or Pirsch Analytics (Germany) - are built primarily for websites, not mobile apps, and lack the mobile SDK depth and product analytics capabilities that PostHog provides. Self-hosting an equivalent stack would require substantial ongoing engineering effort. We treat PostHog as an accepted pragmatic compromise: the data stays in Europe, the data involved is pseudonymous and does not include message content or identifiers linked to your real identity, and we continue to monitor the EU mobile analytics market for a fully sovereign alternative. If a good alternative emerges, we will migrate. The data involved is pseudonymous: user identifiers, device type, in-app events (for example "opened a feature"), and feature flag states. No names, email addresses, or message content are included in analytics. ## Sub-Processors › eustella Subprocessors We want to be honest with you: the mobile app industry was largely built on US infrastructure, and some services that make modern apps work - particularly push notifications and app store billing - are controlled by Apple and Google. This is not unique to eustella; it applies to every app on your phone. Where we have had a genuine choice, we have chosen European providers. Where we have had no choice, we have put legal safeguards in place and minimised the data involved as much as possible. We actively monitor the EU technology landscape and will replace any US-based provider the moment a credible European alternative exists. If you are aware of an EU-headquartered company that could replace any of the US-based providers on this list, please contact us at privacy@eustella.com - we will evaluate it seriously and get back to you. ## Sub-Processors › eustella Subprocessors This list covers companies that process data strictly on our behalf under our instructions. It does not cover the Apple App Store or Google Play Store (which operate as independent controllers under their own privacy policies when you download or make purchases), eustella's own staff and contractors, or the infrastructure sub-suppliers of the companies listed above. ## Sub-Processors › eustella Subprocessors For questions about this list, to be notified of changes, or to suggest a European alternative to any provider listed here: Privacy contact: privacy@eustella.comPrivacy Policy: https://eustella.com/services/privacy This document is reviewed at least annually or upon any material change to our processor landscape. --- # eustella — Service Terms URL: https://eustella.com/services/service-terms Topic: terms ## Service Terms Last updated: 23.06.2026 These Service Terms govern your use of specific eustella features and Services. Capitalised terms not defined here have the meanings given in the eustella Terms of Use ("Agreement"). If there is a conflict between these Service Terms and the Agreement, these Service Terms will prevail. ## Service Terms › eustella Service Terms From time to time we offer features or Services on a beta, early access, preview, or pilot basis ("Beta Services"). Beta Services are clearly identified as such. Nature of Beta Services. Beta Services may contain bugs or errors, may not function as intended, and may be changed, reduced, or withdrawn at any time. Beta Services may involve data loss, resets, or migration as part of development. We will give you reasonable advance notice before any planned reset, migration, or other change that we expect to result in loss of, or material reduced access to, your Content, so that you have an opportunity to export it. No warranties. Beta Services are offered "as-is" and "as-available". To the extent permitted by applicable law, eustella makes no representations or warranties for Beta Services, including any warranty that Beta Services will be generally available, uninterrupted, or error-free, or that Content will be secure or not lost or damaged. Nothing in this section limits or excludes your non-waivable statutory rights as a consumer under EU law, including any rights you may have under Directive (EU) 2019/770 on digital content and digital services. Feedback and improvements. We may invite you to provide feedback on Beta Services. You agree that we may use such feedback to improve, develop, and refine our Services without compensation to you. Providing feedback is voluntary. ## Service Terms › eustella Service Terms eustella may accept images as part of Input ("Image Capabilities"). The following rules apply when you use Image Capabilities: Prohibited uses. You may not use Image Capabilities to attempt to identify a person, to solicit or infer private or sensitive information about a person from their image, or to reproduce the likeness of any person without their express consent and all necessary rights. Your responsibility. You are responsible for ensuring that any image you provide as Input does not violate the rights of any third party, including intellectual property rights, personality rights, and data protection rights under the GDPR. You represent and warrant that you have the necessary rights and permissions for any image you upload. AI-generated images. If eustella may generate images as part of Output, the ownership and usage rules set out in the "Content" section of the Terms of Use apply. You should not represent AI-generated images as photographs or as depicting real events unless clearly disclosed. ## Service Terms › eustella Service Terms eustella allows you to connect third-party applications and services to your account, such as calendars, cloud storage, email, and other productivity tools ("Connected Services"). Authorisation. When you connect a third-party service, you authorise eustella to access and interact with that service on your behalf, within the scope of the permissions you grant. You can review and revoke these permissions at any time in your account settings. Your responsibility. You are responsible for ensuring that your use of Connected Services through eustella complies with the terms and conditions of the relevant third-party service. eustella is not a party to your agreement with any third-party service provider. Data from Connected Services. Data retrieved from Connected Services may be included in the Input or Output of eustella. We process this data solely to provide the Services to you and in accordance with our Privacy Policy. We do not use data from your Connected Services to train AI models. Third-party terms. Connected Services are provided and operated by third parties. eustella does not control and is not responsible for the content, functionality, security, or availability of any Connected Service. If a Connected Service becomes unavailable, changes its terms, or restricts access, eustella's ability to interact with that service may be affected. Security. Do not connect a third-party service unless you trust it and have reviewed its terms and privacy policy. If you believe a Connected Service has been compromised or is being accessed without your authorisation, disconnect it immediately and contact us at security@eustella.com. ## Service Terms › eustella Service Terms eustella can perform tasks and take actions on your behalf, including interacting with Connected Services, drafting and sending communications, modifying files, scheduling events, and — in the future — performing transactions such as making purchases or bookings ("Agentic Capabilities"). You remain in control. Agentic Capabilities act on your instructions and within the permissions you grant. You can configure the level of autonomy the agent has in your account settings. eustella will seek your explicit confirmation before performing actions that are irreversible, financially consequential, or that affect other people, unless you have explicitly configured otherwise. Your responsibility. You are responsible for all actions eustella takes on your behalf. Review the agent's proposed actions before confirming them. eustella does not guarantee that agentic actions will be error-free, timely, or produce the intended result. You should verify outcomes, particularly for actions with legal, financial, or interpersonal consequences. Limitations. Agentic Capabilities depend on the availability and functionality of Connected Services and external systems. eustella cannot guarantee that a third-party service will accept, process, or complete any action initiated by the agent. Safety. For detailed guidance on using Agentic Capabilities safely — including risks of automation, unintended actions, and how to configure appropriate safeguards — please read our Safety Guidelines. ## Service Terms › eustella Service Terms eustella is an AI assistant. It is not a licensed professional and does not replace the judgment of qualified human professionals. The following restrictions apply: Legal services. eustella does not provide legal advice. You may not use eustella to provide tailored legal counsel, prepare legal submissions, represent parties before courts or authorities, or offer legal opinions to third parties in a professional capacity. In Austria, the unauthorised provision of legal services (Winkelschreiberei) is prohibited by law. Similar prohibitions exist across the EU. eustella may provide general legal information for educational purposes, but this does not constitute legal advice and must not be relied upon as such. Medical and health services. eustella is not intended for use in the diagnosis, treatment, or management of any health condition. You may not use eustella as a substitute for professional medical advice, diagnosis, or treatment. Always seek the advice of a qualified healthcare provider with any questions regarding a medical condition. Financial and tax services. eustella does not provide personalised financial, investment, or tax advice. You may not use eustella to provide regulated financial services or tax advisory services to third parties without the involvement of a licensed professional. Other regulated professions. More broadly, you may not use eustella to provide tailored professional advice in any field where such advice requires a licence, registration, or professional qualification under applicable law — including but not limited to accounting, architecture, engineering, pharmacy, and psychotherapy — unless a licensed professional is appropriately involved in reviewing and approving the advice. Disclosure. If you use eustella to assist you in a professional context (for example, to draft a document or research a topic), you remain responsible for the accuracy and appropriateness of the final output, and for complying with any professional or regulatory obligations that apply to you. ## Service Terms › eustella Service Terms No guarantee of accuracy. Output generated by eustella may be incomplete, inaccurate, or out of date. You must evaluate all Output for accuracy and appropriateness before relying on it, sharing it, or using it for any consequential purpose. For more information on AI limitations and risks, see our Safety Guidelines. Third-party content. Output may reference or include content from third-party sources. Such references do not imply endorsement by or affiliation with eustella. Open-source and third-party licences. Where eustella generates code, that code may be similar to or derived from open-source or publicly available code. Such code may be subject to third-party licences, including open-source licences. You are responsible for reviewing and complying with any applicable licence terms. ## Service Terms › eustella Service Terms eustella is not a medical device and is not intended for use in the diagnosis or treatment of any health condition. You are responsible for complying with all applicable laws — including Regulation (EU) 2017/745 on medical devices and any national implementing legislation — for any use of eustella in a medical or healthcare context. Do not use eustella to analyse medical images, diagnostic data, or clinical signals. ## Service Terms › eustella Service Terms We may update these Service Terms from time to time as we develop new features and Services. We will give you at least 30 days' advance notice of changes that materially adversely affect you. New Service Terms sections may be added when new features or Services are launched. © 2026 AI Newsrooms Technology GmbH. All rights reserved. --- # eustella — Sharing & Publication Policy URL: https://eustella.com/services/sharing-and-publication-policy Topic: sharing-policy ## Sharing & Publication Policy Last updated: 23.06.2026 This Sharing & Publication Policy sets out the rules for when you share, publish, or distribute Content generated with or through eustella. It forms part of the eustella Terms of Use. eustella helps you create, draft, research, and act. Much of what you do with eustella will stay private to you. But when you share Output with others - whether by publishing it online, sending it through eustella's agentic features, or using it in a professional or public context - additional responsibilities apply. ## Sharing & Publication Policy › eustella Sharing & Publication Policy You are responsible for all Content you share, publish, send, or distribute using or based on eustella's Output. Before sharing, you should review the Content for accuracy, completeness, and appropriateness. eustella's Output may contain errors, inaccuracies, or omissions. Sharing unverified Output - particularly in contexts where others may rely on it - is done at your own risk. ## Sharing & Publication Policy › eustella Sharing & Publication Policy When you share or publish Content generated or substantially shaped by eustella, you must disclose the involvement of AI in the following situations: Required by law. Under the EU Artificial Intelligence Act (Article 50), you must disclose that content has been artificially generated or manipulated when it constitutes a deep fake (realistic AI-generated images, audio, or video depicting real persons, places, or events) or when AI-generated text is published with the purpose of informing the public on matters of public interest - unless the content has been subject to a human editorial review process where a natural person holds editorial responsibility. These obligations apply from 2 August 2026 at the latest, but eustella encourages early adoption. Required by professional standards. If your profession, institution, or context requires disclosure of AI assistance - for example in academic work, journalism, legal filings, or regulated professional services - you must comply with those requirements. Expected by your audience. Even where not legally required, you should disclose AI involvement where the distinction between human-created and AI-generated content would materially affect how the recipient understands, evaluates, or relies on the content. Transparency builds trust. Artistic and creative exception. Where AI-generated content forms part of an evidently artistic, creative, satirical, or fictional work, the disclosure obligation is limited to making the AI origin known in an appropriate manner that does not hamper the display or enjoyment of the work - for example, in credits, metadata, or an accompanying note. ## Sharing & Publication Policy › eustella Sharing & Publication Policy You may not present eustella's Output as human-created when the distinction matters. Specifically: You may not use eustella to create content that impersonates a real person - including their voice, likeness, or writing style - without that person's explicit consent and in a manner that could mislead others about the content's authenticity. You may not use eustella to generate fake testimonials, reviews, endorsements, news articles, or official documents and present them as authentic. You may not submit eustella's Output as your own original work in contexts where originality is a requirement - including academic assignments, competitive submissions, or applications - unless AI assistance is explicitly permitted and disclosed. ## Sharing & Publication Policy › eustella Sharing & Publication Policy When sharing Content generated by eustella, you must not disclose personal data of other individuals without a lawful basis under the GDPR. This is particularly important when: Content includes personal data retrieved from Connected Services (such as calendar entries, contact details, or files from cloud storage) that relates to people other than you. Content includes information about identifiable individuals that was part of your Input or derived from your conversations with eustella. You use eustella's agentic features to send communications, share documents, or take actions that affect other people. Before eustella sends any content to or about another person, ensure that doing so is lawful, appropriate, and respectful of their privacy. ## Sharing & Publication Policy › eustella Sharing & Publication Policy When eustella acts on your behalf - for example by sending an email, scheduling a meeting, posting a message, or sharing a file - the Content of that action is shared with third parties. The following rules apply: You authorise the sharing. By confirming an agentic action, you instruct eustella to share the relevant Content on your behalf. You are responsible for reviewing the content and the recipient before confirming. Disclosure of AI involvement. When eustella sends a communication on your behalf, the communication should not create the false impression that it was personally written by you if that distinction would materially affect the recipient. Where appropriate, consider including a note that the message was drafted with AI assistance. Third-party terms. Content shared through Connected Services is subject to the terms and conditions of those services. eustella is not responsible for how third-party platforms process, display, or distribute content once it has been sent. ## Sharing & Publication Policy › eustella Sharing & Publication Policy You may not use eustella to generate and share content that infringes the intellectual property rights of others. This includes reproducing substantial portions of copyrighted works, generating content that imitates the distinctive style of a creator in a way that creates confusion about the source, or using trademarks in a misleading manner. Where eustella's Output includes or is based on content from third-party sources (for example, information retrieved through Connected Services or referenced in Output), you are responsible for ensuring that your use and sharing of that content complies with any applicable licences, terms of use, or legal restrictions. ## Sharing & Publication Policy › eustella Sharing & Publication Policy If you share eustella's Output publicly - for example on social media, in a blog post, in a publication, or in a public presentation - the following additional expectations apply: Accuracy. You bear sole responsibility for verifying factual claims before publishing them. eustella's Output should not be treated as a verified source. Attribution. You are not required to attribute content to eustella, but you must not attribute it to another person or source if that attribution would be false or misleading. No endorsement. Sharing Output does not imply that eustella, AI Newsrooms Technology GmbH, or any affiliated entity endorses, verifies, or takes responsibility for the content. ## Sharing & Publication Policy › eustella Sharing & Publication Policy If you use eustella to assist with work in a professional or regulated context - such as legal, medical, financial, journalistic, or academic work - you remain fully responsible for the final output and for complying with all applicable professional and regulatory obligations. See also Section 5 of the eustella Service Terms (Regulated Professional Activities). In particular: In journalism and public-interest publishing, AI-generated text must be disclosed where it informs the public on matters of public interest and has not been subject to effective human editorial control. In academic and educational settings, you must comply with your institution's policies on the use of AI tools. Submitting AI-generated work as your own without disclosure may constitute academic misconduct. In legal and regulatory filings, you are responsible for the accuracy and admissibility of all content. Do not submit AI-generated content to courts, authorities, or regulatory bodies without appropriate review by a qualified professional. ## Sharing & Publication Policy › eustella Sharing & Publication Policy We may remove, restrict, or flag content shared through eustella if we become aware that it violates this Sharing & Publication Policy, the Usage Policies, or applicable law. We may also suspend or terminate the accounts of users who repeatedly or seriously violate this policy. ## Sharing & Publication Policy › eustella Sharing & Publication Policy If you encounter content shared by another eustella user that you believe violates this policy, please report it to security@eustella.com. © 2026 AI Newsrooms Technology GmbH. All rights reserved. --- # eustella — Terms of Use URL: https://eustella.com/services/terms-of-use Topic: terms ## Terms of Use Last updated: 22.04.2026 Thank you for using eustella! These Terms of Use ("Terms") apply to your use of eustella and its associated services, including our mobile applications, web interfaces, AI agent features, and any related software, technology, and websites ("Services"). You form a contract with us when you confirm that you accept these Terms or when you otherwise use the Services. Our Business Terms govern use of eustella for businesses and organisations. Our Privacy Policy explains how we collect and use personal information. Although it does not form part of these Terms, it is an important document that you should read. ## Terms of Use › eustella Terms of Use eustella is built by AI Newsrooms Technology GmbH - a European AI company based in Vienna, Austria. Our mission is to build a trustworthy, sovereign AI agent for Europeans - private, transparent, and independent. For more information, visit eustella.com. AI Newsrooms Technology GmbH Karl-Farkas-Gasse 22, 1030 Vienna, Email: legal@eustella.com Company register: FN 637289s Commercial Court of Vienna ## Terms of Use › eustella Terms of Use Depending on the specific Service or features you use, additional terms and policies may apply. The key ones to be aware of, and which form part of these Terms, are described below: Usage Policy: these policies explain how you may use our Services and Content. Usage PolicyService Terms: these terms apply when you use certain Services or features. eustella Service TermsSharing & Publication Policy: this policy sets out rules for when you share Content. eustella Sharing & Publication PolicyBrand Guidelines: these guidelines explain how you may use our name and logo.Safety Guidelines: these guidelines explain the capabilities and limitations of AI, responsible use, and risks associated with AI-generated content and AI agents. ## Terms of Use › eustella Terms of Use Minimum age. You must be at least 16 years old to use the Services. Registration. You must provide accurate and complete information to register for an account. You may not share your account credentials or make your account available to anyone else and are responsible for all activities that occur under your account. If you create an account on behalf of another person or entity, you must have the authority to accept these Terms on their behalf. Account creation. You may create an account using an email address and password, or by authenticating through a supported third-party identity provider (such as Google or Apple). When you authenticate through a third-party provider, we receive limited profile information (such as your name and email address) in accordance with that provider's terms and our Privacy Policy. ## Terms of Use › eustella Terms of Use What you can do. Subject to your compliance with these Terms, you may access and use our Services. In using our Services, you must comply with all applicable laws as well as the Service-specific terms and policies listed above. What you cannot do. You may not use our Services for any illegal, harmful, or abusive activity. For example, you are prohibited from: Using our Services in a way that infringes, misappropriates, or violates anyone's rights.Misusing our reporting or appeals processes, including providing manifestly unfounded submissions.Modifying, copying, leasing, selling, or distributing any of our Services.Attempting to or assisting anyone to reverse engineer, decompile, or discover the source code or underlying components of our Services, including our models, algorithms, or systems (except to the extent this restriction is prohibited by applicable law).Automatically or programmatically extracting data or Output (defined below).Representing that Output was made purely by a human when it was not.Interfering with or disrupting our Services, including circumventing any rate limits or restrictions or bypassing any protective measures or safety mitigations we put on our Services.Using our Services or Output to develop AI models that compete with eustella.Using agentic features to perform actions you are not authorised to perform, or to circumvent access controls of any connected third-party service. Agentic features. Our Services may include agentic capabilities that can perform tasks and take actions on your behalf, including interacting with third-party services you have connected (such as calendars, cloud storage, or other applications). When you enable agentic features, you authorise the proactive action of your eustella agent to act on your behalf within the scope of the permissions you grant. You remain responsible for all actions taken by the agent on your behalf and should review the agent's actions regularly. Third-party services. Our Services allow you to connect third-party applications, platforms, and services ("Third-Party Services"). When you connect a Third-Party Service, you authorise eustella to access and interact with that service on your behalf, within the scope of the permissions you grant. Third-Party Services are subject to their own terms and privacy policies. We are not responsible for the content, functionality, security, or practices of any Third-Party Service you connected to eustella. You may disconnect any Third-Party Service at any time through your account settings. Software. Our Services may allow you to download software, such as mobile applications, which may update automatically to ensure you're using the latest version. Our software may include open-source software that is governed by its own licences that we have made available to you. Feedback. We appreciate your feedback about our Services, but you agree that we may use it to provide, maintain, develop, and improve our Services, without compensation to you. ## Terms of Use › eustella Terms of Use Your content. You may provide input to the Services ("Input"), and receive output from the Services based on the Input ("Output"). Input and Output are collectively "Content". You are responsible for Content, including ensuring that it does not violate any applicable law or these Terms. You represent and warrant that you have all rights, licences, and permissions needed to provide Input to our Services. Ownership of content. As between you and eustella, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any and possible by law, in and to Output. Similarity of content. Due to the nature of artificial intelligence, Output may not be original and other users may receive similar output from our Services. Our assignment above does not extend to other users' output or any Third-Party Output. Our use of content. We may use your Content to provide, maintain, and improve our Services, to comply with applicable law, to enforce our terms and policies, and to keep our Services safe. We use your Content solely to operate the Services for you. No model training on your data. We do not use your Content - including your conversations, documents, files, or any data from connected Third-Party Services - to train or fine-tune AI models. Your data is yours and stays yours. In the future, we may allow an option to voluntarily opt-in to model training, if you or other users want to help create the next generation of European AI Models. Content moderation. We use automated systems and human review to identify and address content that violates our Terms, including our Usage Policies. If we become aware of violations, we may remove or restrict content, limit your access to features, or suspend or terminate your account. eustella acts as a hosting service provider under the Digital Services Act (DSA). To comply with our DSA obligations regarding transparency and procedural fairness, we provide dedicated mechanisms for reporting illegal content and contesting moderation decisions. If you identify content you believe is harmful or illegal, you may report it through our in-app reporting tool or by emailing support@eustella.com. Where we take significant moderation actions, we will provide a "Statement of Reasons". You have the right to appeal such decision by contacting support@eustella.com. You also have the right to engage in out-of-court dispute settlement through a certified and independent dispute settlement body. Accuracy. Artificial intelligence and machine learning are rapidly evolving fields. We are constantly working to improve our Services to make them more accurate, reliable, safe, and beneficial. Given the probabilistic nature of machine learning, use of our Services may in some situations result in Output that does not accurately reflect real people, places, or facts. When you use our Services you understand and agree: Output may not always be accurate. You should not rely on Output from our Services as a sole source of truth or factual information, or as a substitute for professional advice.You must evaluate Output for accuracy and appropriateness for your use case, including using human review as appropriate, before using or sharing Output from the Services.You must not use any Output relating to a person for any purpose that could have a legal or material impact on that person, such as making credit, educational, employment, housing, insurance, legal, medical, or other important decisions about them.Our Services may provide incomplete, incorrect, or offensive Output that does not represent eustella's views. In that case users may report this behavior which will allow us to increase the safety and performance of our Services. If Output references any third-party products or services, it does not mean the third party endorses or is affiliated with eustella.When using agentic features that take real-world actions (such as sending messages, modifying files, or interacting with connected services), you should exercise particular care and review the agent's proposed actions before confirming them. ## Terms of Use › eustella Terms of Use We and our affiliates own all rights, title, and interest in and to the Services. You may only use our name and logo in accordance with our Brand Guidelines. ## Terms of Use › eustella Terms of Use Paid subscriptions. Some of our Services allow you to purchase paid subscriptions to benefit from enhanced features and functionalities (the "Benefits"). The exact nature of the Benefits differs between Services and will be made clear to you before purchase. You can manage your paid subscription from your account settings. Fees. All charges, including subscription fees, will be made clear to you before purchase. All prices are inclusive of applicable VAT unless stated otherwise. Billing. If you sign up for a paid subscription or purchase any Services, you will provide complete and accurate billing information, including a valid payment method. For paid subscriptions, we will automatically charge your payment method on each agreed-upon periodic renewal until you cancel. If your payment cannot be completed, we may downgrade your account or suspend your access to our Services until payment is received. Right of withdrawal (cooling-off period). If you are a consumer habitually resident in the EU or EEA and you purchase a paid subscription, you have the right to withdraw from the contract within 14 days from the day on which the contract was concluded, without giving any reason and without incurring any costs other than those described below. To exercise your right of withdrawal, you must inform us of your decision to withdraw by an unequivocal statement before the 14-day period has expired. You may use the model withdrawal form set out in the Annex at the end of these Terms, but doing so is not obligatory — any clear statement is sufficient, including an email to legal@eustella.com, an in-product withdrawal request, or a letter sent to our address above. To meet the withdrawal deadline, it is enough for you to send your communication concerning the exercise of your right of withdrawal before the 14-day period expires. If, at the moment of purchase, you expressly request that we begin providing the paid Services during the 14-day cooling-off period and you acknowledge that you will lose your right of withdrawal once the Services have been fully performed, then: if you withdraw before the Services have been fully performed, you must pay an amount which is in proportion to the part of the Services already provided up to the time you communicated your withdrawal, calculated on the basis of the total subscription price agreed in the contract; andonce the Services have been fully performed within the 14-day period (which, for a periodic subscription, means the end of that subscription period), you lose the right of withdrawal in respect of that performance, in accordance with Article 16(a) of Directive 2011/83/EU. If you do not make that express request and acknowledgment, you remain entitled to a full refund within the 14-day period, regardless of any usage during it. Refund. If you withdraw from the contract within the 14-day period, we will reimburse you all payments received from you, less any proportionate amount as described above, without undue delay and in any event no later than 14 days after the day on which we are informed of your withdrawal. We will use the same means of payment as you used for the initial transaction unless you have expressly agreed otherwise; in any event, you will not incur any fees as a result of the reimbursement. Cancellation outside the cooling-off period. Following the 14-day cooling-off period, you can cancel your paid subscription at any time by updating your account settings. You will not be charged after you cancel. You will continue to have access to the Benefits until the end of the subscription period you have paid for, at which point your cancellation will become effective. Unless we specify otherwise or unless required by your statutory rights, you will not receive a refund or service credit for any days between the day you cancel and the last day of the subscription period you have paid for. Price changes. We may change our prices from time to time. If we increase our subscription prices, we will give you at least 30 days' notice and any price increase will take effect on your next renewal so that you can cancel if you do not agree to the price increase. ## Terms of Use › eustella Terms of Use Your rights. You can stop using our Services and end your relationship with eustella at any time by closing your account and stopping your use of the Services. Instructions for how to do this are available in your account settings. If you are an EU-based consumer who has just accepted these Terms, your 14-day right of withdrawal is described in the "Right of withdrawal (cooling-off period)" section above. Our rights. We may take action to restrict, suspend, or terminate your access to our Services or close your account if we determine, acting reasonably and objectively: You breached these Terms, including our Usage Policies.We must do so to comply with the law.Your use of our Services could cause risk or harm to eustella, our users, or anyone else.Your account has been inactive for more than 12 months without an active paid subscription, to uphold our data minimization and security principles. We recommend regularly exporting any Content you wish to preserve. Appeals. If you believe we have suspended or terminated your account in error, you can file an appeal by contacting support@eustella.com. ## Terms of Use › eustella Terms of Use How we provide the Services. We commit to provide the Services to you with reasonable skill and care and to act with professional diligence. We do not promise to offer the Services forever or in their current form for any particular period of time. Liability. Provided that we have acted with professional diligence, we do not take responsibility for loss or damage caused by us, unless it is caused by our breach of these Terms or reasonably foreseeable at the time of entering into these Terms. We do not take responsibility for loss or damage caused by events beyond our reasonable control. We do not exclude or limit our liability to you in any way where it would be unlawful for us to do so. You still have the full protections of the laws applicable to you. Your statutory rights. You have certain statutory rights as an European Citizen that cannot be limited or excluded by these Terms. These Terms are in no way intended to affect or restrict those rights. Some, but not all of those rights, include: the right to withdraw from a distance contract within 14 days without giving any reason (EU Consumer Rights Directive);protection against unfair contract terms that create a significant imbalance to your detriment (EU Unfair Contract Terms Directive, Austrian Consumer Protection Act / KSchG);protection against unfair, misleading, or aggressive commercial practices (EU Unfair Commercial Practices Directive);the right to access, rectify, erase, restrict, and port your personal data, and to object to its processing (EU General Data Protection Regulation);the right to clear, comprehensible pre-contractual information before you are bound by the contract (EU Consumer Rights Directive); andany additional rights provided under the laws of your country of residence that apply to you as a consumer. This list is provided for your convenience, does not constitute legal advice and might not be exhaustive. In case of any conflict between these Terms and your mandatory statutory rights, your statutory rights prevail. Consumer guarantee. EU consumer laws provide you with a legal guarantee covering the Services. If you have questions about your legal guarantee, please contact support@eustella.com or look at our FAQ section. ## Terms of Use › eustella Terms of Use Informal resolution first. If we have a dispute, we would first like to understand and try to address your concerns before any formal proceedings. Please contact us at legal@eustella.com. We will make good-faith efforts to resolve the matter within 30 days of receiving your notice. Governing law. These Terms and any dispute or claim arising out of or in connection with these Terms or the Services (including non-contractual disputes or claims, and including their subject matter or formation) are governed by and construed in accordance with the laws of Austria, excluding its conflict-of-law rules. If you are a consumer habitually resident in an EU or EEA Member State, you additionally enjoy the protection of any mandatory provisions of the law of your country of residence. Nothing in these Terms affects your rights as a consumer under the mandatory laws that apply to you. Users within the European Economic Area. If you are a consumer habitually resident in an EEA Member State, you may bring proceedings against us in the courts of your country of residence, and we may only bring proceedings against you in the courts of the Member State in which you are domiciled. This is required by the Brussels Ia Regulation (EU) No. 1215/2012 and cannot be overridden by these Terms. If you are not a consumer (for example, you use eustella for business or professional purposes) and are established within the EEA, you agree that the courts of Vienna, Austria, shall have exclusive jurisdiction. Users outside the European Economic Area. If you reside outside the European Economic Area, the following hierarchy applies to any dispute between you and eustella that cannot be resolved informally: (1) Default — mandatory individual arbitration. Any such dispute, claim, or controversy arising out of or in connection with these Terms or the Services shall be resolved exclusively through binding individual arbitration, rather than in court. This agreement to arbitrate is governed by the Federal Arbitration Act (9 U.S.C. §§ 1 et seq.) to the extent you reside in the United States, and by Austrian arbitration law (§§ 577 et seq. ZPO) for all other users outside the EEA. Arbitration shall be administered by the Vienna International Arbitral Centre (VIAC) under its Rules of Arbitration in effect at the time the arbitration is commenced. The seat of arbitration shall be Vienna, Austria. The language of the arbitration shall be English. The arbitration shall be conducted by a sole arbitrator. (2) Fallback — Vienna courts. The courts of Vienna, Austria, shall have exclusive jurisdiction only in the limited cases where (a) you have validly opted out of arbitration under the procedure described below, or (b) a particular claim is severed from arbitration because the class action waiver below is found unenforceable as to that claim. Class action waiver. If you reside outside the European Economic Area, you agree that any dispute resolution proceedings will be conducted only on an individual basis and not as part of a class, consolidated, or representative action. You waive any right to participate in a class action lawsuit or class-wide arbitration against us. If this waiver is found to be unenforceable with respect to a particular claim, that claim — and only that claim — shall be severed and may proceed in court, while all other claims remain subject to individual arbitration. 30-day opt-out. If you reside outside the EEA, you may opt out of the mandatory arbitration and class action waiver provisions by sending written notice to legal@eustella.com within 30 days of first accepting these Terms. Your notice must include your name, the email address associated with your eustella account, and a clear statement that you wish to opt out. If you opt out, the Vienna courts referred to in subsection (2) above will have exclusive jurisdiction over any dispute or claim arising out of or in connection with these Terms or the Services. Alternative dispute resolution for EEA users. If you reside in the EEA, you may also raise a dispute with an alternative dispute resolution body. The Internet Ombudsperson (Österreichisches Institut für angewandte Telekommunikation) is the competent out-of-court dispute resolution body in Austria for disputes arising from contracts entered into online. ## Terms of Use › eustella Terms of Use Assignment. You may not assign or transfer any rights or obligations under these Terms without our prior written consent. We may assign or transfer our rights or obligations under these Terms (a) by operation of law in connection with a merger or demerger of AI Newsrooms Technology GmbH (universal succession), or (b) in connection with a sale of all or substantially all of our assets or the business associated with the Services to a successor in interest. If you are a consumer, we will not transfer our obligations or this contract to a not-individually-named third party with debt-discharging effect without your express consent, in line with § 6 Abs 2 Z 2 of the Austrian Consumer Protection Act (KSchG) and equivalent mandatory provisions of EU consumer protection law. In any case, your statutory rights as a consumer will not be affected by any assignment. Changes to these Services. We are continuously working to maintain and improve our Services. We may make modifications to the Services that are necessary to keep the Services in conformity with these Terms (for example, security updates, bug fixes, and performance improvements) at any time and at no additional cost to you. Beyond such conformity-maintaining changes, we may also modify the Services where we have a valid reason to do so — including, for example, to adapt the Services to a new technical environment, to changes in the number of users, or to new legal or regulatory requirements; to add, replace, or discontinue features; or to align the Services with our product roadmap. Any such modification will be made at no extra cost to you, and we will inform you in a clear and comprehensible manner. If a modification negatively affects your access to or use of the Services in more than a minor way, we will give you reasonable advance notice on a durable medium (for example, by email) of the features and timing of the modification, and of your right to terminate the contract free of charge within 30 days of receiving the notice or of the modification taking effect, whichever is later. This is in line with Article 19 of Directive (EU) 2019/770. If you have purchased a paid subscription and you terminate on this basis, we will reimburse the unused portion of any pre-paid fees on a pro-rata basis. Changes to these Terms. We may update these Terms from time to time. We will give you at least 30 days' advance notice of any material change, either via email or an in-product notification. Material changes will only apply going forward. If you do not agree to a material change, you may stop using the Services and terminate your account before the changes take effect; if you continue to use the Services after the effective date, you accept the updated Terms. Trade controls. You must comply with all applicable trade laws, including sanctions and export control laws. ## Terms of Use › eustella Terms of Use If any provision of these Terms is found to be invalid, unlawful, or unenforceable by a competent court or authority, that provision shall be deemed severed from these Terms. The remaining provisions shall continue in full force and effect, provided that the Terms can reasonably continue to operate without the severed provision. Where a provision is found to be unfair within the meaning of Directive 93/13/EEC on unfair terms in consumer contracts (as implemented by the Austrian Consumer Protection Act / KSchG), it shall not be binding on you as a consumer. We will not seek to replace it with a substitute term that approximates our original intent. Any gap resulting from the removal of an unfair term shall be governed by the applicable statutory provisions of Austrian law. If the removal of any provision would make these Terms as a whole incapable of being performed or eustella cannot reasonably provide the services under certain provisions, both parties may terminate the contractual relationship by giving written notice to the other. ## Terms of Use › eustella Terms of Use (Complete and return this form only if you wish to withdraw from the contract.) To: AI Newsrooms Technology GmbH, Karl-Farkas-Gasse 22, 1030 Vienna, Austria, legal@eustella.com Withdrawal Notice I hereby give notice that I withdraw from my contract for the provision of the following service: Contract details Date of conclusion of the contract: Consumer information Name of consumer: Address of consumer: Date and signature of consumer: © 2026 AI Newsrooms Technology GmbH. All rights reserved. --- # eustella — Trust Center URL: https://eustella.com/services/trust Topic: trust ## Trust Center Last updated: 23.06.2026 This Trust Center is the place to go when something feels off while using eustella, when you need to understand how your data is handled, or when you want to see what safety and privacy commitments we actually stand behind. It is grouped by the kind of problem you might have. If you came here with a specific concern, jump to the section that matches it. If you just want the overview, start at the top. ## Trust Center › eustella Trust Center We do not train AI models on your data. Your conversations, files, preferences, and data from any services you connect are not used to train, fine-tune, or improve AI models. We do not sell your data. We do not rent or trade your personal data, and eustella does not run advertising. GDPR rights apply in full. You can access, correct, export, delete, or restrict your data at any time, and we respond within one month. We use open-source AI models we run ourselves. This means we can tell you which model generated a response, where it ran, and what it has access to. Closed beta. eustella is currently in a closed beta testing phase. Access is invitation-only. Your privacy rights and safety commitments apply in full during the beta - see Service Terms for beta-specific conditions. ## Trust Center › eustella Trust Center AI systems are not perfect. eustella will sometimes give you answers that are wrong, misleading, biased, or inappropriate. Here is how to recognise the common failure modes and what to do about each one. Hallucinations - eustella made something up. Large language models generate text that sounds plausible even when the underlying facts are wrong. eustella can invent quotes, citations, dates, statistics, names, laws, or entire events. Treat any factual claim from eustella as a starting point to verify, not as a source of truth - especially for medical, legal, financial, safety-critical, or news-related questions. If you spot a hallucination, delete or regenerate the response, and report it to us if it is about an identifiable person or caused real harm. See the AI Safety Guide for a full treatment. Sycophancy - eustella is agreeing with you too much. LLMs tend to tell users what they want to hear. If eustella has changed its mind as soon as you pushed back, or validated a plan you suspect is flawed, take its agreement with a grain of salt. Ask it to argue the opposite position or to list the strongest objections. If you need an honest second opinion, do not rely on eustella alone. Bias or stereotyping. Models trained on internet-scale data reflect the biases present in that data. If eustella produces a response that stereotypes a group, reinforces prejudice, or treats a protected characteristic unfairly, please report it to support@eustella.com with the full prompt and response. Do not just regenerate and move on - we can only fix what we see. Inaccurate personal information about you. If eustella has saved an incorrect fact about you (in personalisation) or generated incorrect content about you within a conversation, you have the right to correct it under Art. 16 GDPR. You can edit saved preferences directly in the app, delete the conversation, or email privacy@eustella.com to request a formal rectification. Inaccurate information about another person. If eustella generates false or defamatory content about a third party, do not share or publish it. If you believe you have already shared a response containing such content, delete it where possible and email support@eustella.com. Unsafe, harmful, or illegal output. If eustella produces content that violates our Usage Policy - for example content that could facilitate self-harm, violence, child sexual abuse material, or illegal activity - stop using that conversation immediately and report it to support@eustella.com. Include the prompt, the response, and any context. Minors safety reports can also go to privacy@eustella.com. Jailbreak or prompt-injection attempts. If you see a response where eustella appears to follow instructions hidden in a document, web page, or file you uploaded - instead of your instructions - that may be a prompt injection. Report it to security@eustella.com so we can look at the safety pipeline. You are not sure which of the above applies. Send it to support@eustella.com and we will route it. ## Trust Center › eustella Trust Center If something has gone wrong with your data, or you want to exercise a right you have under the GDPR. Privacy Policy (app) - What data we collect, why, on what legal basis, and for how long.Sub-processors - Who processes data on our behalf.Deletion Requests - How to request full or partial deletion of your account and data.Website Privacy Policy - The separate privacy notice for eustella.com. Typical situations and where to go. You want to delete your account or specific data - see Deletion Requests or email privacy@eustella.com.You want a copy of everything we hold about you - email privacy@eustella.com. This is your right to access under Art. 15 GDPR. We respond within one month.eustella has an incorrect preference or memory about you - you can edit or clear it directly in the app, or email privacy@eustella.com.You suspect your account has been accessed by someone else - email security@eustella.com immediately and change your password if you can still sign in.You believe a minor under 16 is using eustella - email privacy@eustella.com. We will investigate and delete the account and data.You want to complain about how we handle data - you can always email us at privacy@eustella.com. You also have the right to lodge a complaint with your national data protection authority (in Austria: Österreichische Datenschutzbehörde, dsb@dsb.gv.at). ## Trust Center › eustella Trust Center If you want to understand what is actually generating the responses you see, or the principles we apply when choosing models. AI Models - The models that power eustella, the role each one plays, and why we selected them.AI Safety Guide - Hallucinations, sycophancy, best practices, and your rights when AI gets things wrong.Digital Sovereignty - Why we run open-source models on our own systems rather than routing requests to third-party APIs. ## Trust Center › eustella Trust Center The rules that govern using eustella, what we commit to, and what is not allowed. Terms of Use - The agreement between you and eustella.Service Terms - Additional terms for specific features, including the current closed beta.Usage Policy - What you may and may not do with eustella.Sharing and Publication Policy - Rules for sharing or publishing content generated with eustella. ## Trust Center › eustella Trust Center Imprint - Legal information about AI Newsrooms Technology GmbH, the company behind eustella.Press & Brand Guidelines - Company boilerplate, logos, and contact for journalists and partners. ## Trust Center › eustella Trust Center A quick reference for which address to use when something is wrong. SituationWhere to reportHallucination, bias, unsafe output, content moderation issuesupport@eustella.comGDPR request (access, rectification, deletion, portability, objection)privacy@eustella.comAccount takeover, suspected breach, prompt injection, vulnerabilitysecurity@eustella.comA minor under 16 is using eustellaprivacy@eustella.comLegal claims, compliance, regulatory questionslegal@eustella.com © 2026 AI Newsrooms Technology GmbH. All rights reserved. --- # eustella — Nutzungslimits URL: https://eustella.com/services/usage-limits/de Topic: services Last updated: 2026-07-06 ## Nutzungslimits Stand: 06.07.2026 Diese Seite erklärt, wie die Nutzungslimits bei eustella funktionieren, damit Sie vor Abschluss eines Abonnements wissen, was Sie erwartet. Sie dient der Transparenz. Auf dieser Seite geht es darum, wie viel Sie eustella nutzen können. Die Regeln dazu, wie Sie eustella nutzen dürfen — was erlaubt ist und was nicht —, finden Sie in der Nutzungsrichtlinie. ## Nutzungslimits › Nutzungslimits & faire Nutzung bei eustella eustella erfasst die Nutzung als ein einziges, gemeinsames wöchentliches Kontingent. Chats, Bildgenerierung, Websuche, das Lesen von Dateien und Agent-Läufe schöpfen alle aus demselben Topf, sodass Sie Ihr Kontingent ganz nach Bedarf einsetzen können, statt an getrennte Quoten je Funktion gebunden zu sein. Als grobe Orientierung: Ein Free-Tarif reicht für rund 200 Chats pro Monat.* Diese Angaben sind indikative Entsprechungen desselben Kontingents, keine eigenständigen Quoten — nutzen Sie das eine, verringert sich das andere, denn beide teilen sich einen Topf. Bezahlte Tarife bieten Ihnen dasselbe Kontingent, nur vervielfacht — so können Sie mehr von beidem tun, in beliebiger Mischung: TarifWöchentliches KontingentUngefähr pro Monat (beliebige Mischung)Free1×~200 Chats / 20 BilderComet~4×~800 Chats / 80 BilderStar~12×~2.400 Chats / 240 BilderCosmos~60×~12.000 Chats / 1.200 Bilder Was in den einzelnen Tarifen enthalten ist, sehen Sie unter Preise. Möchten Sie mehr aus jedem Chat herausholen? Erfahren Sie in unserem Best-Practices-Leitfaden, wie Sie Ihr Kontingent optimal nutzen. * Ein Hinweis zu diesen Zahlen. Die oben genannten Chat- und Bildzahlen sind eine grobe Näherung — nicht die tatsächlich mögliche Anzahl an Chats, die je nach Anwendungsfall stark variiert. eustella hat sie mit unseren Best Practices im Hinterkopf berechnet, ohne sogenannte Tool-Calls (Websuchen, Dokumentenerstellung, Speichereinträge und Ähnliches) und mit einer einfachen Soul-Einstellung von eustella. Reale Nutzung mit Tool-Calls, längeren Gesprächen oder einer umfangreicheren Soul verbraucht Ihr Kontingent schneller — betrachten Sie diese Zahlen daher als grobe Orientierung und nicht als Zusage. ## Nutzungslimits › Nutzungslimits & faire Nutzung bei eustella eustella erfasst Ihre Nutzung anhand dieser Elemente: Wöchentliches Kontingent — Ihr Hauptkontingent. Das ist das oben beschriebene Kontingent. Es wird jede Woche zurückgesetzt.Rollierendes 5-Stunden-Limit — ein kurzfristiger Schutz (Star und Cosmos). In den Tarifen Star und Cosmos steht neben Ihrem wöchentlichen Kontingent ein kürzeres, über die letzten 5 Stunden gemessenes rollierendes Limit, damit ein einzelner intensiver Schub es nicht auf einen Schlag aufbraucht. Andere Tarife haben dieses Limit nicht.Aufladeguthaben — ein optionaler Überlauftopf. Wenn Sie Guthaben aufladen, wird es erst dann genutzt, wenn Ihr reguläres Kontingent aufgebraucht ist — niemals vorher —, und es wird übertragen, statt zurückgesetzt zu werden. Wenn Sie Ihr Limit erreichen, pausieren einige Funktionen, bis sich Ihr Kontingent zurücksetzt; Ihre Daten und Ihr Verlauf bleiben davon unberührt. In der App wird Ihre Nutzung als Prozentsatz Ihres aktuellen Kontingents angezeigt und nicht als laufende Einheitenzählung — die obigen Zahlen sind indikative Entsprechungen genau dieses Kontingents. Manche Konten sind ohne Limit (unbegrenzt); bei diesen Konten wird keine Nutzungsanzeige eingeblendet. ## Nutzungslimits › Nutzungslimits & faire Nutzung bei eustella › So funktioniert das Kontingent Alles, was eustella tut, läuft über das zugrunde liegende KI-Modell, und genau dafür zahlen Sie mit Ihrem Kontingent. Zwei Arten von Aktivität zehren daran. Denken und Schreiben. Jede Antwort bedeutet, dass eustella Ihre Anfrage durchdenkt und eine Antwort formuliert — das sind die Grundkosten jeder Chat-Nachricht. Eine kurze Frage kostet wenig; lange, komplexe Antworten und lange Unterhaltungen kosten mehr, ganz einfach, weil das Modell mehr lesen und erzeugen muss. Tool-Aufrufe. Um zu handeln und nicht nur zu reden, nutzt eustella Tools — das Web durchsuchen, eine Datei oder eine Erinnerung lesen, ein Bild ansehen, ein Bild erzeugen und mehr. Je nach Tool kostet ein Aufruf entweder ungefähr so viel, wie es braucht, um die zurückgelieferten Informationen zu lesen (dieser Text gelangt wie jede andere Eingabe in den Chat), oder er kostet mehr, wenn das Tool selbst teuer im Betrieb ist. Im Einzelnen: Etwas zu lesen kostet ungefähr so viel wie es einzufügen. Wenn eustella eine Erinnerung liest, eine Datei öffnet oder ein von Ihnen gesendetes Bild ansieht, wird dieser Inhalt Teil dessen, was das Modell aufnehmen muss — es kostet also ungefähr so viel, wie wenn Sie diesen Text oder dieses Dokument selbst in den Chat kopiert hätten. Eine kurze Notiz ist günstig; ein langer Bericht kostet mehr. Nichts Verstecktes — schlicht die Länge dessen, was gelesen wird.Websuche kostet in zweifacher Hinsicht. Erstens berechnet uns der europäische Anbieter, dessen Suchindex wir nutzen, für jede Suche einen kleinen Betrag. Zweitens zählen die zurückgelesenen Ergebnisse als gewöhnliche Eingabe — genau so, als wäre dieser Text in Ihrem Chat eingegangen. Eine schnelle Abfrage ist geringfügig; eine Sitzung voller Suchen summiert sich.Bildgenerierung kostet mehr. Ein Bild zu erstellen bedeutet, ein größeres, teureres Bildgenerierungsmodell aufzurufen, sodass ein erzeugtes Bild spürbar mehr von Ihrem Kontingent verbraucht als eine schriftliche Antwort. Gut zu wissen, bevor Sie ein Dutzend Varianten erzeugen.Agenten kosten so viel, wie dieselbe Arbeit von Hand kosten würde. Ein Agent schöpft aus demselben wöchentlichen Kontingent wie Ihre Chats: jede Suche, die er durchführt, jedes Dokument, das er liest, jedes Bild, das er erzeugt, und der Bericht, den er schreibt, zählen alle — genau wie in einem normalen Chat. Zu beachten ist vor allem, dass sich ein Zeitplan wiederholt — ein täglicher Recherche-Agent verbraucht still und leise jeden Tag den Gegenwert einer vollständigen Recherchesitzung, ob Sie das Ergebnis lesen oder nicht. Planen Sie daher bewusst: Wählen Sie den Rhythmus, den Sie tatsächlich nutzen, nicht den häufigsten möglichen. Erreicht ein Agent Ihr Limit, wird sein Lauf einfach pausiert und im nächsten Zeitfenster erneut versucht, und Ihr Zeitplan bleibt bestehen. Der 5-Stunden-Schutz in den Tarifen Star und Cosmos (siehe oben) gilt weiterhin. Im Gegensatz dazu ist eine einfache Chat-Nachricht ungefähr das Günstigste, was Sie tun können — ein weiterer Grund, warum kurze, fokussierte Chats Ihr Kontingent weiter reichen lassen. ## Nutzungslimits › Nutzungslimits & faire Nutzung bei eustella Wenn Sie ein kostenpflichtiges Abonnement abschließen, zunächst einmal: Danke, dass Sie uns unterstützen. Ein Abonnement verschafft Ihnen mehr Nutzungskontingent — einen größeren gemeinsamen Topf, den Sie über Chats, Suchen, Bilder und Agenten hinweg einsetzen können, und damit mehr Spielraum, um mehr mit eustella zu tun. Was mehr Kontingent nicht verschafft, sind perfekte, fehlerfreie Antworten. Wie jede KI heute kann eustella in jedem Tarif Fehler machen oder halluzinieren — ein größeres Kontingent gibt Ihnen mehr Spielraum zum Arbeiten, Suchen und Ausführen von Agenten, ist aber keine Garantie für Richtigkeit. Wie Sie die besten und zuverlässigsten Ergebnisse aus eustella herausholen, erfahren Sie in unserem Leitfaden: So nutzen Sie eustella — Best Practices. ## Nutzungslimits › Nutzungslimits & faire Nutzung bei eustella Für Geschäfts- und API-Kunden sind Nutzungslimits, Rate-Limits und Quoten in Ihrem Tarif oder in einer individuellen schriftlichen Vereinbarung festgelegt, wie in den Geschäftsbedingungen beschrieben. API-Rate-Limits sind für Entwickler separat dokumentiert. ## Nutzungslimits › Nutzungslimits & faire Nutzung bei eustella eustella kann diese Kontingente im Laufe der Zeit anpassen, wenn sich Modelle, Kosten und Nutzungsmuster weiterentwickeln. ## Nutzungslimits › Nutzungslimits & faire Nutzung bei eustella Leistungsstärkere Modelloptionen. eustella führt möglicherweise Modelloptionen mit höherer Leistungsfähigkeit ein. Sollte dies geschehen, verbrauchen solche Modelle pro Aktion mehr von Ihrem Kontingent als das Standardmodell, sodass dasselbe Kontingent für weniger Aktionen mit leistungsstarken Modellen reicht. Die obigen Entsprechungen sind auf Basis des Standardmodells geschätzt. ## Nutzungslimits › Nutzungslimits & faire Nutzung bei eustella Wenn Sie eine Frage zu Ihrem Kontingent haben oder der Ansicht sind, dass ein Limit fälschlicherweise angewendet wurde, kontaktieren Sie uns unter support@eustella.com. © 2026 AI Newsrooms Technology GmbH. Alle Rechte vorbehalten. --- # eustella — Límites de uso URL: https://eustella.com/services/usage-limits/es Topic: services Last updated: 2026-07-06 ## Límites de uso Última actualización: 06.07.2026 Esta página explica cómo funcionan los límites de uso en eustella para que sepas qué esperar antes de suscribirte. La ofrecemos por transparencia. Esta página trata sobre cuánto puedes usar eustella. Para conocer las reglas sobre cómo puedes usarla —qué está permitido y qué no— consulta la Política de uso. ## Límites de uso › Límites de uso y uso razonable de eustella eustella mide el uso mediante una asignación semanal única y compartida. Los chats, la generación de imágenes, la búsqueda web, la lectura de archivos y las ejecuciones de agentes se nutren todos del mismo conjunto, de modo que puedes repartir tu asignación como mejor te convenga en lugar de quedar atado a cuotas independientes por función. Como guía aproximada, un plan Free basta para unos 200 chats al mes.* Estas cifras son equivalencias indicativas de la misma asignación, no cuotas independientes: usar una reduce la otra, porque comparten un solo conjunto. Los planes de pago te ofrecen la misma asignación, multiplicada, para que puedas hacer más de cualquiera de las dos, en la proporción que quieras: PlanAsignación semanalAprox. al mes (cualquier mezcla)Free1×~200 chats / 20 imágenesComet~4×~800 chats / 80 imágenesStar~12×~2.400 chats / 240 imágenesCosmos~60×~12.000 chats / 1.200 imágenes Consulta los Precios para ver qué incluye cada plan. ¿Quieres sacar más partido a cada chat? Descubre cómo maximizar tu asignación en nuestra guía de buenas prácticas. * Una nota sobre estas cifras. Los recuentos de chats e imágenes anteriores son una estimación aproximada — no el número real de chats posibles, que varía mucho según el caso de uso. eustella los calculó teniendo en cuenta nuestras buenas prácticas, sin ninguna de las llamadas «tool calls» (búsquedas web, creación de documentos, escrituras en memoria y similares) y con una configuración de Soul básica de eustella. El uso real que implica llamadas a herramientas, conversaciones más largas o una Soul más rica consume tu asignación más rápido, así que trata estas cifras como una orientación aproximada y no como una promesa. ## Límites de uso › Límites de uso y uso razonable de eustella eustella mide tu uso con estos elementos: Asignación semanal — tu asignación principal. Es la asignación descrita más arriba. Se restablece cada semana.Límite móvil de 5 horas — una salvaguarda a corto plazo (Star y Cosmos). En los planes Star y Cosmos, un límite móvil más corto medido sobre las últimas 5 horas convive con tu asignación semanal, para que una ráfaga intensa no la agote de una sola sentada. Los demás planes no tienen este límite.Crédito adicional — un fondo opcional para excedentes. Si añades crédito, se utiliza solo después de agotar tu asignación habitual —nunca antes— y se acumula en lugar de restablecerse. Si alcanzas tu límite, algunas funciones se pausan hasta que tu asignación se restablece; tus datos y tu historial no se ven afectados. En la app, tu uso se muestra como un porcentaje de tu asignación actual en lugar de un recuento de unidades acumuladas: las cifras anteriores son equivalencias indicativas de esa misma asignación. Algunas cuentas son sin medición (sin límite); esas cuentas no ven ninguna barra de uso. ## Límites de uso › Límites de uso y uso razonable de eustella › Cómo funciona la asignación Todo lo que hace eustella se ejecuta sobre el modelo de IA subyacente, y tu asignación es lo que lo paga. Dos tipos de actividad la reducen. Pensar y escribir. Cada respuesta implica que eustella razone tu solicitud y redacte una respuesta: este es el coste básico de cualquier mensaje de chat. Una pregunta corta cuesta poco; las respuestas largas y elaboradas y las conversaciones extensas cuestan más, sencillamente porque hay más que leer y que producir para el modelo. Llamadas a herramientas. Para actuar y no solo hablar, eustella utiliza herramientas: buscar en la web, leer un archivo o una memoria, mirar una imagen, generar una imagen y más. Según la herramienta, una llamada cuesta aproximadamente lo que cuesta leer la información que devuelve (ese texto entra en el chat como cualquier otra entrada), o cuesta más cuando la propia herramienta es cara de ejecutar. En detalle: Leer algo cuesta más o menos lo que costaría pegarlo. Cuando eustella lee una memoria, abre un archivo o mira una imagen que envías, ese contenido pasa a formar parte de lo que el modelo debe asimilar, así que cuesta aproximadamente lo que costaría si tú mismo hubieras copiado y pegado ese texto o documento en el chat. Una nota breve es barata; un informe largo cuesta más. Nada oculto: simplemente la extensión de lo que se lee.La búsqueda web cuesta por dos motivos. Primero, el proveedor europeo cuyo índice de búsqueda utilizamos nos cobra una pequeña cantidad por cada búsqueda. Segundo, los resultados que se leen cuentan como entrada ordinaria, exactamente como si ese texto hubiera llegado a tu chat. Una consulta rápida es mínima; una sesión llena de búsquedas se acumula.La generación de imágenes cuesta más. Crear una imagen implica llamar a un modelo de generación de imágenes más grande y más caro, por lo que una imagen generada consume bastante más de tu asignación que una respuesta escrita. Conviene saberlo antes de generar una docena de variaciones.Los agentes cuestan lo que costaría hacer ese mismo trabajo a mano. Un agente se nutre de la misma asignación semanal que tus chats: cada búsqueda que hace, cada documento que lee, cada imagen que crea y el informe que escribe cuentan todos, igual que en un chat normal. Lo único a tener en cuenta es que una programación se repite: un agente de investigación diario gasta discretamente el equivalente a una sesión de investigación completa cada día, leas o no el resultado. Así que programa con criterio: elige la frecuencia que realmente vayas a usar, no la más alta disponible. Si un agente alcanza tu límite, su ejecución simplemente se pausa y se reintenta en la siguiente ventana, y tu programación se mantiene intacta. La salvaguarda de 5 horas en los planes Star y Cosmos (más arriba) sigue aplicándose. En cambio, un simple mensaje de chat es prácticamente lo más barato que puedes hacer: otra razón por la que los chats breves y centrados hacen que tu asignación rinda más. ## Límites de uso › Límites de uso y uso razonable de eustella Si te suscribes a un plan de pago, antes que nada: gracias por apoyarnos. Una suscripción te da más asignación de uso: un conjunto compartido más grande para gastar en chats, búsquedas, imágenes y agentes, y por tanto más margen para hacer más con eustella. Lo que más asignación no te da son respuestas perfectas y sin errores. Como toda IA hoy en día, eustella puede cometer errores o alucinar en cualquier plan; una asignación mayor te da más margen para trabajar, buscar y ejecutar agentes, pero no es una garantía de exactitud. Para saber cómo obtener los mejores y más fiables resultados de eustella, consulta nuestra guía: Cómo usar eustella — buenas prácticas. ## Límites de uso › Límites de uso y uso razonable de eustella Para clientes empresariales y de API, los límites de uso, los límites de frecuencia y las cuotas se establecen en tu plan o en un acuerdo individual por escrito, tal como se describe en los Términos para empresas. Los límites de frecuencia de la API se documentan por separado para desarrolladores. ## Límites de uso › Límites de uso y uso razonable de eustella eustella puede ajustar estas asignaciones con el tiempo a medida que evolucionan los modelos, los costes y los patrones de uso. ## Límites de uso › Límites de uso y uso razonable de eustella Opciones de modelos más potentes. eustella podría introducir opciones de modelos de mayor capacidad. Si lo hace, esos modelos consumen más asignación por acción que el modelo estándar, por lo que la misma asignación alcanza para menos acciones con modelos pesados. Las equivalencias anteriores se calculan con el modelo estándar. ## Límites de uso › Límites de uso y uso razonable de eustella Si tienes alguna duda sobre tu asignación o crees que se ha aplicado un límite por error, escríbenos a support@eustella.com. © 2026 AI Newsrooms Technology GmbH. Todos los derechos reservados. --- # eustella — Limites d'utilisation URL: https://eustella.com/services/usage-limits/fr Topic: services Last updated: 2026-07-06 ## Limites d'utilisation Dernière mise à jour : 06.07.2026 Cette page explique le fonctionnement des limites d'utilisation sur eustella, afin que vous sachiez à quoi vous attendre avant de vous abonner. Elle est fournie dans un souci de transparence. Cette page traite de la quantité d'utilisation d'eustella dont vous disposez. Pour les règles sur la manière dont vous pouvez l'utiliser — ce qui est autorisé et ce qui ne l'est pas — consultez la Politique d'utilisation. ## Limites d'utilisation › Limites d'utilisation et usage équitable d'eustella eustella mesure l'utilisation sous la forme d'une enveloppe hebdomadaire unique et partagée. Les conversations, la génération d'images, la recherche web, la lecture de fichiers et les exécutions d'agents puisent toutes dans le même réservoir : vous répartissez ainsi votre enveloppe comme bon vous semble, sans être enfermé dans des quotas distincts par fonctionnalité. À titre indicatif, un forfait Free suffit pour environ 200 conversations par mois.* Ces chiffres sont des équivalents indicatifs d'une même enveloppe, et non des quotas indépendants : utiliser l'un réduit l'autre, puisqu'ils partagent un seul réservoir. Les forfaits payants vous offrent la même enveloppe, multipliée — vous pouvez donc faire davantage de l'un ou de l'autre, dans n'importe quelle proportion : ForfaitEnveloppe hebdomadaireEnviron par mois (toutes proportions)Free1×~200 conversations / 20 imagesComet~4×~800 conversations / 80 imagesStar~12×~2 400 conversations / 240 imagesCosmos~60×~12 000 conversations / 1 200 images Consultez la page Tarifs pour savoir ce qu'inclut chaque forfait. Envie de tirer le meilleur de chaque conversation ? Découvrez comment optimiser votre enveloppe dans notre guide des bonnes pratiques. * À propos de ces chiffres. Les nombres de conversations et d'images ci-dessus sont une approximation grossière — et non le nombre réel de conversations possibles, qui varie beaucoup selon l'usage. eustella les a calculés en gardant à l'esprit nos bonnes pratiques, sans aucun appel d'outil (« tool call » : recherches web, création de documents, écritures en mémoire, etc.) et avec un réglage de Soul basique pour eustella. Un usage réel impliquant des appels d'outils, des conversations plus longues ou une Soul plus riche épuise votre enveloppe plus vite : considérez donc ces chiffres comme une orientation approximative, pas comme une promesse. ## Limites d'utilisation › Limites d'utilisation et usage équitable d'eustella eustella mesure votre utilisation à l'aide des éléments suivants : Enveloppe hebdomadaire — votre enveloppe principale. Il s'agit de l'enveloppe décrite ci-dessus. Elle se réinitialise chaque semaine.Limite glissante sur 5 heures — un garde-fou à court terme (Star et Cosmos). Sur les forfaits Star et Cosmos, une limite glissante plus courte, mesurée sur les 5 dernières heures, vient s'ajouter à votre enveloppe hebdomadaire, de sorte qu'un pic d'activité intense ne puisse pas la consommer en une seule séance. Les autres forfaits n'ont pas cette limite.Crédit de recharge — un réservoir de dépassement facultatif. Si vous ajoutez du crédit, il n'est utilisé qu'une fois votre enveloppe habituelle épuisée — jamais avant — et il se reporte au lieu de se réinitialiser. Si vous atteignez votre limite, certaines fonctionnalités sont suspendues jusqu'à la réinitialisation de votre enveloppe ; vos données et votre historique ne sont pas affectés. Dans l'application, votre utilisation est affichée sous forme de pourcentage de votre enveloppe actuelle plutôt que sous forme de décompte d'unités — les chiffres ci-dessus sont des équivalents indicatifs de cette même enveloppe. Certains comptes sont sans limite (non mesurés) ; ces comptes ne voient aucune barre d'utilisation. ## Limites d'utilisation › Limites d'utilisation et usage équitable d'eustella › Comment fonctionne l'enveloppe Tout ce que fait eustella repose sur le modèle d'IA sous-jacent, et c'est précisément ce que votre enveloppe finance. Deux types d'activité l'entament. Réfléchir et rédiger. Chaque réponse suppose qu'eustella analyse votre demande et rédige une réponse — c'est le coût de base de tout message de conversation. Une question brève coûte peu ; les réponses longues et élaborées ainsi que les conversations qui s'étirent coûtent davantage, tout simplement parce que le modèle a plus à lire et à produire. Appels d'outils. Pour agir et non seulement discuter, eustella utilise des outils — rechercher sur le web, lire un fichier ou une mémoire, regarder une image, générer une image, et plus encore. Selon l'outil, un appel coûte soit à peu près ce qu'il faut pour lire l'information renvoyée (ce texte entre dans la conversation comme n'importe quelle autre saisie), soit davantage lorsque l'outil lui-même est coûteux à exécuter. Dans le détail : Lire quelque chose coûte à peu près autant que de le coller. Lorsqu'eustella lit une mémoire, ouvre un fichier ou regarde une image que vous envoyez, ce contenu fait désormais partie de ce que le modèle doit assimiler — cela coûte donc à peu près ce que cela coûterait si vous aviez vous-même copié-collé ce texte ou ce document dans la conversation. Une note brève est peu coûteuse ; un long rapport coûte davantage. Rien de caché — simplement la longueur de ce qui est lu.La recherche web coûte à deux titres. D'abord, le fournisseur européen dont nous utilisons l'index de recherche nous facture une petite somme pour chaque recherche. Ensuite, les résultats relus comptent comme une saisie ordinaire — exactement comme si ce texte était arrivé dans votre conversation. Une recherche rapide est négligeable ; une séance remplie de recherches finit par peser.La génération d'images coûte davantage. Créer une image implique de faire appel à un modèle de génération d'images plus grand et plus coûteux : une image générée consomme donc nettement plus de votre enveloppe qu'une réponse écrite. Bon à savoir avant de produire une douzaine de variantes.Les agents coûtent ce que coûterait le même travail effectué à la main. Un agent puise dans la même enveloppe hebdomadaire que vos conversations : chaque recherche qu'il effectue, chaque document qu'il lit, chaque image qu'il crée et le rapport qu'il rédige comptent tous, exactement comme dans une conversation normale. Le seul point de vigilance est qu'une planification se répète — un agent de recherche quotidien dépense discrètement l'équivalent d'une séance de recherche complète chaque jour, que vous en lisiez le résultat ou non. Planifiez donc à bon escient : choisissez la fréquence que vous utiliserez réellement, et non la plus élevée possible. Si un agent atteint votre limite, son exécution est simplement suspendue et relancée à la fenêtre suivante, et votre planification reste intacte. Le garde-fou de 5 heures des forfaits Star et Cosmos (ci-dessus) s'applique toujours. À l'inverse, un simple message de conversation est à peu près la chose la moins coûteuse que vous puissiez faire — une raison de plus pour laquelle des conversations brèves et ciblées font durer votre enveloppe. ## Limites d'utilisation › Limites d'utilisation et usage équitable d'eustella Si vous souscrivez à un forfait payant, tout d'abord : merci de nous soutenir. Un abonnement vous procure davantage d'enveloppe d'utilisation — un réservoir partagé plus grand à répartir entre conversations, recherches, images et agents, et donc plus de marge pour faire davantage avec eustella. Ce que davantage d'enveloppe ne vous procure pas, ce sont des réponses parfaites et sans erreur. Comme toute IA aujourd'hui, eustella peut encore se tromper ou halluciner, quel que soit le forfait — une enveloppe plus grande vous laisse plus de marge pour travailler, rechercher et exécuter des agents, mais ce n'est pas une garantie d'exactitude. Pour savoir comment obtenir les résultats les plus fiables et les plus aboutis avec eustella, consultez notre guide : Comment utiliser eustella — bonnes pratiques. ## Limites d'utilisation › Limites d'utilisation et usage équitable d'eustella Pour les clients professionnels et API, les limites d'utilisation, les limites de débit et les quotas sont définis dans votre forfait ou dans un accord écrit individuel, comme décrit dans les Conditions professionnelles. Les limites de débit de l'API font l'objet d'une documentation distincte destinée aux développeurs. ## Limites d'utilisation › Limites d'utilisation et usage équitable d'eustella eustella peut ajuster ces enveloppes au fil du temps, à mesure qu'évoluent les modèles, les coûts et les habitudes d'utilisation. ## Limites d'utilisation › Limites d'utilisation et usage équitable d'eustella Des options de modèles plus puissants. eustella pourra proposer des options de modèles aux capacités supérieures. Le cas échéant, ces modèles consomment une plus grande part de votre enveloppe par action que le modèle standard : la même enveloppe couvre donc un moins grand nombre d'actions avec les modèles plus gourmands. Les équivalents ci-dessus sont estimés pour le modèle standard. ## Limites d'utilisation › Limites d'utilisation et usage équitable d'eustella Si vous avez une question sur votre enveloppe ou si vous pensez qu'une limite a été appliquée par erreur, écrivez-nous à support@eustella.com. © 2026 AI Newsrooms Technology GmbH. Tous droits réservés. --- # eustella — Usage Limits URL: https://eustella.com/services/usage-limits Topic: services Last updated: 2026-07-06 ## Usage Limits Last updated: 06.07.2026 This page explains how usage limits work on eustella so you know what to expect before you subscribe. It is provided for transparency. This page is about how much you can use eustella. For the rules on how you may use it — what is and isn't allowed — see the Usage Policy. ## Usage Limits › eustella Usage Limits & Fair Use eustella meters usage as a single, shared weekly allowance. Chats, image generation, web search, file reading, and agent runs all draw from the same pool, so you can spend your allowance however suits you rather than being locked into separate per-feature quotas. As a rough guide, one Free plan is enough for about 200 chats per month.* These figures are indicative equivalents of the same allowance, not independent quotas — using one draws down the other, because they share one pool. Paid plans give you the same allowance, multiplied — so you can do more of either, in any mix: PlanWeekly allowanceRoughly per month (any mix)Free1×~200 chats / 20 imagesComet~4×~800 chats / 80 imagesStar~12×~2,400 chats / 240 imagesCosmos~60×~12,000 chats / 1,200 images See Pricing for what each plan includes. Want to get more out of every chat? Find out how to maximise your usage in our best practices guide. * A note on these figures. The chat and image counts above are a gross approximation — not the actual number of chats possible, which varies a lot by use case. eustella calculated them with our best practices in mind, without any so-called tool calls (web searches, document creation, memory writes, and the like) and with a basic Soul setting for eustella. Real usage that involves tool calls, longer conversations, or a richer Soul draws down your allowance faster, so treat these numbers as a rough orientation rather than a promise. ## Usage Limits › eustella Usage Limits & Fair Use eustella meters your usage with these elements: Weekly allowance — your main allowance. This is the allowance described above. It resets each week.5-hour rolling limit — a short-term guard (Star and Cosmos). On the Star and Cosmos plans, a shorter rolling limit measured over the last 5 hours sits alongside your weekly allowance, so one intense burst can't consume it in a single sitting. Other plans do not have this limit.Top-up credit — an optional overflow pool. If you add credit, it is used only after your regular allowance is full — never before — and it carries over rather than resetting. If you reach your limit, some capabilities pause until your allowance resets; your data and history are unaffected. In the app, your usage is shown as a percentage of your current allowance rather than a running unit count — the figures above are indicative equivalents of that same allowance. Some accounts are unmetered (no limit); those accounts see no usage bar. ## Usage Limits › eustella Usage Limits & Fair Use › How the allowance works Everything eustella does runs on the underlying AI model, and your allowance is what pays for it. Two kinds of activity draw it down. Thinking and writing. Every reply means eustella reasoning through your request and writing an answer — this is the baseline cost of any chat message. A short question costs little; long, involved answers and long conversations cost more, simply because there is more for the model to read and to produce. Tool calls. To act rather than only talk, eustella uses tools — searching the web, reading a file or a memory, looking at an image, generating a picture, and more. Depending on the tool, a call either costs about what it takes to read the information it returns (that text enters the chat like any other input), or it costs more when the tool itself is expensive to run. In detail: Reading something costs about what pasting it would. When eustella reads a memory, opens a file, or looks at an image you send, that content becomes part of what the model has to take in — so it costs roughly what it would if you had copy-pasted that text or document into the chat yourself. A short note is cheap; a long report costs more. Nothing hidden — simply the length of what is read.Web search costs on two counts. First, the European provider we use for the search index charges us a little for every search. Second, the results read back count as ordinary input — exactly as if that text had arrived in your chat. One quick lookup is minor; a session full of searches adds up.Image generation costs more. Creating an image means calling a larger, more expensive image-generation model, so a generated picture uses noticeably more of your allowance than a written reply. Worth knowing before you produce a dozen variations.Agents cost what the same work would cost by hand. An agent draws from the same weekly allowance as your chats: every search it makes, every document it reads, every image it creates, and the report it writes all count, just as they would in a normal chat. The one thing to watch is that a schedule repeats — a daily research agent quietly spends a full research session's worth of allowance every day, whether or not you read the result. So schedule deliberately: choose the cadence you will actually use, not the most frequent one available. If an agent reaches your limit, its run simply pauses and retries in the next window, and your schedule stays intact. The 5-hour guard on the Star and Cosmos plans (above) still applies. By contrast, a plain chat message is about the cheapest thing you can do — another reason short, focused chats stretch your allowance further. ## Usage Limits › eustella Usage Limits & Fair Use If you subscribe to a paid plan, first of all: thank you for supporting us. A subscription buys you more usage allowance — a larger shared pool to spend across chats, searches, images, and agents, and so more room to do more with eustella. What more allowance does not buy is perfect, error-free answers. Like every AI today, eustella can still make mistakes or hallucinate on any plan — a bigger allowance gives you more room to work, search & run agents, but it is not a guarantee of accuracy. For how to get the best and most reliable results from eustella, see our guide: How to use eustella — best practices. ## Usage Limits › eustella Usage Limits & Fair Use For business and API customers, usage limits, rate limits, and quotas are set out in your plan or in an individual written agreement, as described in the Business Terms. API rate limits are documented separately for developers. ## Usage Limits › eustella Usage Limits & Fair Use eustella may adjust these allowances over time as models, costs, and usage patterns evolve. ## Usage Limits › eustella Usage Limits & Fair Use More powerful model options. eustella may introduce higher-capability model options. If it does, those models consume more of your allowance per action than the standard model, so the same allowance stretches to fewer heavy-model actions. The equivalents above are estimated at the standard model. ## Usage Limits › eustella Usage Limits & Fair Use If you have a question about your allowance or believe a limit has been applied in error, contact us at support@eustella.com. © 2026 AI Newsrooms Technology GmbH. All rights reserved. --- # eustella — Limiti di utilizzo URL: https://eustella.com/services/usage-limits/it Topic: services Last updated: 2026-07-06 ## Limiti di utilizzo Ultimo aggiornamento: 06.07.2026 Questa pagina spiega come funzionano i limiti di utilizzo di eustella, così sai cosa aspettarti prima di sottoscrivere un abbonamento. Ha uno scopo puramente informativo, per garantire la massima trasparenza. Questa pagina riguarda quanto puoi usare eustella. Per le regole su come puoi usarla — ciò che è consentito e ciò che non lo è — consulta le Norme di utilizzo. ## Limiti di utilizzo › Limiti di utilizzo e uso corretto di eustella eustella misura l'utilizzo come un unico plafond settimanale condiviso. Le chat, la generazione di immagini, la ricerca web, la lettura di file e le esecuzioni degli agenti attingono tutte allo stesso pool, così puoi distribuire il tuo plafond come preferisci, senza essere vincolato a quote separate per ogni funzione. Come regola di massima, un piano Free basta per circa 200 chat al mese.* Questi valori sono equivalenti indicativi dello stesso plafond, non quote indipendenti: usarne uno riduce l'altro, perché condividono un unico pool. I piani a pagamento offrono lo stesso plafond, moltiplicato — così puoi fare di più con entrambi, nella combinazione che preferisci: PianoPlafond settimanaleCirca al mese (in qualsiasi combinazione)Free1×~200 chat / 20 immaginiComet~4×~800 chat / 80 immaginiStar~12×~2.400 chat / 240 immaginiCosmos~60×~12.000 chat / 1.200 immagini Consulta i Prezzi per scoprire cosa include ciascun piano. Vuoi ottenere di più da ogni chat? Scopri come massimizzare il tuo plafond nella nostra guida alle buone pratiche. * Una nota su questi numeri. I conteggi di chat e immagini qui sopra sono un'approssimazione di massima — non il numero reale di chat possibili, che varia molto a seconda dell'uso. eustella li ha calcolati tenendo presenti le nostre buone pratiche, senza alcuna cosiddetta «tool call» (ricerche sul web, creazione di documenti, scritture in memoria e simili) e con un'impostazione di Soul di base di eustella. Un uso reale che comporta chiamate a strumenti, conversazioni più lunghe o una Soul più ricca consuma il tuo plafond più rapidamente: considera quindi questi numeri come un'indicazione di massima e non come una promessa. ## Limiti di utilizzo › Limiti di utilizzo e uso corretto di eustella eustella misura il tuo utilizzo attraverso questi elementi: Plafond settimanale — il tuo plafond principale. È il plafond descritto sopra. Si azzera ogni settimana.Limite mobile su 5 ore — una tutela a breve termine (Star e Cosmos). Sui piani Star e Cosmos, un limite mobile più breve, calcolato sulle ultime 5 ore, affianca il plafond settimanale, in modo che un singolo picco intenso non lo esaurisca in un'unica sessione. Gli altri piani non hanno questo limite.Credito di ricarica — un pool aggiuntivo facoltativo. Se aggiungi credito, viene utilizzato solo dopo che il plafond ordinario è esaurito — mai prima — e viene riportato invece di azzerarsi. Se raggiungi il tuo limite, alcune funzionalità si sospendono fino all'azzeramento del plafond; i tuoi dati e la tua cronologia restano intatti. Nell'app, il tuo utilizzo è mostrato come percentuale del plafond attuale anziché come conteggio progressivo di unità — i valori sopra indicati sono equivalenti indicativi dello stesso plafond. Alcuni account sono senza limiti (nessun limite); questi account non vedono alcuna barra di utilizzo. ## Limiti di utilizzo › Limiti di utilizzo e uso corretto di eustella › Come funziona il plafond Tutto ciò che fa eustella si basa sul modello di IA sottostante, ed è proprio questo che il tuo plafond paga. Due tipi di attività lo riducono. Pensare e scrivere. Ogni risposta comporta che eustella ragioni sulla tua richiesta e scriva una risposta: è il costo di base di qualsiasi messaggio in chat. Una domanda breve costa poco; risposte lunghe e articolate e conversazioni prolungate costano di più, semplicemente perché il modello ha di più da leggere e da produrre. Chiamate agli strumenti. Per agire e non solo parlare, eustella usa degli strumenti: cercare sul web, leggere un file o una memoria, guardare un'immagine, generare un'immagine e altro ancora. A seconda dello strumento, una chiamata costa all'incirca quanto serve per leggere le informazioni che restituisce (quel testo entra nella chat come qualsiasi altro input), oppure costa di più quando lo strumento stesso è oneroso da eseguire. Nel dettaglio: Leggere qualcosa costa più o meno quanto incollarlo. Quando eustella legge una memoria, apre un file o guarda un'immagine che invii, quel contenuto entra a far parte di ciò che il modello deve assimilare, quindi costa all'incirca quanto costerebbe se avessi copiato e incollato tu stesso quel testo o documento nella chat. Una nota breve costa poco; un rapporto lungo costa di più. Nulla di nascosto: semplicemente la lunghezza di ciò che viene letto.La ricerca web costa su due fronti. Primo, il fornitore europeo di cui usiamo l'indice di ricerca ci addebita un piccolo importo per ogni ricerca. Secondo, i risultati riletti contano come input ordinario, esattamente come se quel testo fosse arrivato nella tua chat. Una rapida consultazione è minima; una sessione piena di ricerche si somma.La generazione di immagini costa di più. Creare un'immagine significa chiamare un modello di generazione di immagini più grande e più costoso, quindi un'immagine generata consuma sensibilmente più plafond di una risposta scritta. Utile saperlo prima di produrre una dozzina di varianti.Gli agenti costano quanto costerebbe fare lo stesso lavoro a mano. Un agente attinge allo stesso plafond settimanale delle tue chat: ogni ricerca che effettua, ogni documento che legge, ogni immagine che crea e il rapporto che scrive contano tutti, proprio come in una chat normale. L'unica cosa a cui fare attenzione è che una pianificazione si ripete: un agente di ricerca quotidiano spende con discrezione l'equivalente di un'intera sessione di ricerca ogni giorno, che tu ne legga o meno il risultato. Pianifica quindi con criterio: scegli la frequenza che userai davvero, non la più alta disponibile. Se un agente raggiunge il tuo limite, la sua esecuzione viene semplicemente sospesa e ritentata nella finestra successiva, e la tua pianificazione resta intatta. La tutela di 5 ore dei piani Star e Cosmos (sopra) continua ad applicarsi. Al contrario, un semplice messaggio in chat è più o meno la cosa più economica che puoi fare: un altro motivo per cui chat brevi e mirate fanno durare di più il tuo plafond. ## Limiti di utilizzo › Limiti di utilizzo e uso corretto di eustella Se sottoscrivi un piano a pagamento, prima di tutto: grazie per il tuo sostegno. Un abbonamento ti dà più plafond di utilizzo — un pool condiviso più ampio da spendere tra chat, ricerche, immagini e agenti, e quindi più margine per fare di più con eustella. Ciò che più plafond non ti dà sono risposte perfette e prive di errori. Come ogni IA di oggi, eustella può comunque sbagliare o avere allucinazioni con qualsiasi piano: un plafond più ampio ti dà più margine per lavorare, cercare ed eseguire agenti, ma non è una garanzia di accuratezza. Per scoprire come ottenere i risultati migliori e più affidabili da eustella, consulta la nostra guida: Come usare eustella — buone pratiche. ## Limiti di utilizzo › Limiti di utilizzo e uso corretto di eustella Per i clienti business e API, i limiti di utilizzo, i limiti di frequenza e le quote sono stabiliti nel tuo piano o in un accordo scritto individuale, come descritto nelle Condizioni per le aziende. I limiti di frequenza delle API sono documentati separatamente per gli sviluppatori. ## Limiti di utilizzo › Limiti di utilizzo e uso corretto di eustella eustella può modificare questi plafond nel tempo, con l'evolversi dei modelli, dei costi e delle modalità di utilizzo. ## Limiti di utilizzo › Limiti di utilizzo e uso corretto di eustella Opzioni di modelli più potenti. eustella potrebbe introdurre opzioni di modelli con capacità superiori. In tal caso, questi modelli consumano una parte maggiore del tuo plafond per ogni azione rispetto al modello standard, quindi lo stesso plafond consente meno azioni con i modelli più impegnativi. Gli equivalenti sopra indicati sono stimati sul modello standard. ## Limiti di utilizzo › Limiti di utilizzo e uso corretto di eustella Se hai una domanda sul tuo plafond o ritieni che un limite sia stato applicato per errore, scrivici all'indirizzo support@eustella.com. © 2026 AI Newsrooms Technology GmbH. Tutti i diritti riservati. --- # eustella — Gebruikslimieten URL: https://eustella.com/services/usage-limits/nl Topic: services Last updated: 2026-07-06 ## Gebruikslimieten Laatst bijgewerkt: 06.07.2026 Deze pagina legt uit hoe gebruikslimieten bij eustella werken, zodat je weet wat je kunt verwachten voordat je een abonnement neemt. We bieden dit voor de transparantie. Deze pagina gaat over hoeveel je eustella kunt gebruiken. Voor de regels over hoe je het mag gebruiken — wat wel en niet is toegestaan — zie het Gebruiksbeleid. ## Gebruikslimieten › Gebruikslimieten en redelijk gebruik van eustella eustella meet gebruik als één enkel, gedeeld weektegoed. Chats, het genereren van afbeeldingen, webzoekopdrachten, het lezen van bestanden en agent-runs putten allemaal uit dezelfde pool, dus je kunt je tegoed besteden zoals het jou uitkomt in plaats van vast te zitten aan aparte quota per functie. Als vuistregel is één Free-abonnement genoeg voor ongeveer 200 chats per maand.* Deze cijfers zijn indicatieve equivalenten van hetzelfde tegoed, geen losstaande quota — als je het ene gebruikt, gaat dat ten koste van het andere, omdat ze één pool delen. Betaalde abonnementen geven je hetzelfde tegoed, maar vermenigvuldigd — zodat je meer van beide kunt doen, in elke gewenste verhouding: AbonnementWeektegoedOngeveer per maand (in elke mix)Free1×~200 chats / 20 afbeeldingenComet~4×~800 chats / 80 afbeeldingenStar~12×~2.400 chats / 240 afbeeldingenCosmos~60×~12.000 chats / 1.200 afbeeldingen Zie Prijzen voor wat elk abonnement omvat. Wil je meer uit elke chat halen? Ontdek hoe je je tegoed maximaliseert in onze best practices-gids. * Een opmerking over deze cijfers. De bovenstaande aantallen chats en afbeeldingen zijn een ruwe benadering — niet het werkelijke aantal mogelijke chats, dat sterk verschilt per gebruikssituatie. eustella heeft ze berekend met onze best practices in gedachten, zonder zogeheten tool calls (webzoekopdrachten, het maken van documenten, geheugenschrijfacties en dergelijke) en met een basisinstelling voor de Soul van eustella. Werkelijk gebruik met tool calls, langere gesprekken of een rijkere Soul verbruikt je tegoed sneller — beschouw deze cijfers dus als een ruwe richtlijn en niet als een belofte. ## Gebruikslimieten › Gebruikslimieten en redelijk gebruik van eustella eustella meet je gebruik met deze onderdelen: Weektegoed — je belangrijkste tegoed. Dit is het tegoed dat hierboven is beschreven. Het wordt elke week gereset.Voortschrijdende limiet van 5 uur — een kortetermijnbeveiliging (Star en Cosmos). Bij de abonnementen Star en Cosmos geldt naast je weektegoed een kortere voortschrijdende limiet, gemeten over de laatste 5 uur, zodat één intensieve piek het niet in één keer kan opmaken. Andere abonnementen hebben deze limiet niet.Aanvullend tegoed — een optionele overloop-pool. Als je tegoed toevoegt, wordt dit pas gebruikt nadat je reguliere tegoed op is — nooit eerder — en het schuift door in plaats van te resetten. Als je je limiet bereikt, worden sommige functies gepauzeerd totdat je tegoed wordt gereset; je gegevens en geschiedenis blijven onaangetast. In de app wordt je gebruik weergegeven als een percentage van je huidige tegoed in plaats van als een lopende eenhedenteller — de cijfers hierboven zijn indicatieve equivalenten van datzelfde tegoed. Sommige accounts zijn ongelimiteerd (geen limiet); die accounts zien geen gebruiksbalk. ## Gebruikslimieten › Gebruikslimieten en redelijk gebruik van eustella › Hoe het tegoed werkt Alles wat eustella doet, draait op het onderliggende AI-model, en dat is precies waar je tegoed voor betaalt. Twee soorten activiteit putten eruit. Denken en schrijven. Elk antwoord betekent dat eustella je verzoek doordenkt en een antwoord opstelt — dat zijn de basiskosten van elk chatbericht. Een korte vraag kost weinig; lange, uitgebreide antwoorden en lange gesprekken kosten meer, simpelweg omdat het model meer moet lezen en produceren. Tool-aanroepen. Om te handelen en niet alleen te praten, gebruikt eustella tools — op het web zoeken, een bestand of een herinnering lezen, naar een afbeelding kijken, een afbeelding genereren en meer. Afhankelijk van de tool kost een aanroep ongeveer evenveel als het lezen van de informatie die hij teruggeeft (die tekst komt in de chat terecht als elke andere invoer), of kost hij meer wanneer de tool zelf duur is om uit te voeren. In detail: Iets lezen kost ongeveer evenveel als het plakken ervan. Wanneer eustella een herinnering leest, een bestand opent of naar een afbeelding kijkt die je stuurt, wordt die inhoud onderdeel van wat het model moet opnemen — het kost dus ongeveer evenveel als wanneer je die tekst of dat document zelf in de chat had geplakt. Een korte notitie is goedkoop; een lang rapport kost meer. Niets verborgens — simpelweg de lengte van wat er wordt gelezen.Webzoekopdrachten kosten op twee manieren. Ten eerste rekent de Europese aanbieder wiens zoekindex we gebruiken ons een klein bedrag voor elke zoekopdracht. Ten tweede tellen de teruggelezen resultaten als gewone invoer — precies alsof die tekst in je chat was binnengekomen. Eén snelle opzoeking is minimaal; een sessie vol zoekopdrachten telt op.Het genereren van afbeeldingen kost meer. Een afbeelding maken betekent dat er een groter, duurder beeldgeneratiemodel wordt aangeroepen, dus een gegenereerde afbeelding verbruikt merkbaar meer van je tegoed dan een geschreven antwoord. Goed om te weten voordat je een dozijn varianten maakt.Agenten kosten wat hetzelfde werk met de hand zou kosten. Een agent put uit hetzelfde weektegoed als je chats: elke zoekopdracht die hij uitvoert, elk document dat hij leest, elke afbeelding die hij maakt en het rapport dat hij schrijft tellen allemaal mee, net als in een gewone chat. Het enige om op te letten is dat een planning zich herhaalt — een dagelijkse onderzoeksagent besteedt stilletjes elke dag het equivalent van een volledige onderzoekssessie, of je het resultaat nu leest of niet. Plan dus bewust: kies de frequentie die je daadwerkelijk gaat gebruiken, niet de hoogst mogelijke. Als een agent je limiet bereikt, wordt zijn run simpelweg gepauzeerd en in het volgende venster opnieuw geprobeerd, en blijft je planning intact. De beveiliging van 5 uur bij de abonnementen Star en Cosmos (hierboven) blijft van toepassing. Ter vergelijking: een gewoon chatbericht is ongeveer het goedkoopste dat je kunt doen — nog een reden waarom korte, gerichte chats je tegoed verder laten reiken. ## Gebruikslimieten › Gebruikslimieten en redelijk gebruik van eustella Als je een betaald abonnement neemt, allereerst: bedankt dat je ons steunt. Een abonnement geeft je meer gebruikstegoed — een grotere gedeelde pool om te besteden aan chats, zoekopdrachten, afbeeldingen en agenten, en dus meer ruimte om meer met eustella te doen. Wat meer tegoed je niet geeft, zijn perfecte, foutloze antwoorden. Zoals elke AI van vandaag kan eustella nog steeds fouten maken of hallucineren, op welk abonnement dan ook — een groter tegoed geeft je meer ruimte om te werken, te zoeken en agenten uit te voeren, maar het is geen garantie voor juistheid. Lees in onze gids hoe je de beste en meest betrouwbare resultaten uit eustella haalt: Hoe je eustella gebruikt — best practices. ## Gebruikslimieten › Gebruikslimieten en redelijk gebruik van eustella Voor zakelijke klanten en API-klanten worden gebruikslimieten, rate limits en quota vastgelegd in je abonnement of in een individuele schriftelijke overeenkomst, zoals beschreven in de Zakelijke voorwaarden. API rate limits worden apart gedocumenteerd voor ontwikkelaars. ## Gebruikslimieten › Gebruikslimieten en redelijk gebruik van eustella eustella kan deze tegoeden na verloop van tijd aanpassen naarmate modellen, kosten en gebruikspatronen veranderen. ## Gebruikslimieten › Gebruikslimieten en redelijk gebruik van eustella Krachtigere modelopties. eustella kan modelopties met hogere capaciteit introduceren. Als dat gebeurt, verbruiken die modellen per actie meer van je tegoed dan het standaardmodel, waardoor hetzelfde tegoed toereikend is voor minder acties met een zwaar model. De equivalenten hierboven zijn geschat op basis van het standaardmodel. ## Gebruikslimieten › Gebruikslimieten en redelijk gebruik van eustella Als je een vraag hebt over je tegoed of denkt dat een limiet ten onrechte is toegepast, neem dan contact met ons op via support@eustella.com. © 2026 AI Newsrooms Technology GmbH. Alle rechten voorbehouden. --- # eustella — Limity użycia URL: https://eustella.com/services/usage-limits/pl Topic: services Last updated: 2026-07-06 ## Limity użycia Ostatnia aktualizacja: 06.07.2026 Ta strona wyjaśnia, jak działają limity użycia w eustella, abyś wiedział, czego się spodziewać przed wykupieniem subskrypcji. Publikujemy ją w trosce o przejrzystość. Ta strona dotyczy tego, ile możesz korzystać z eustella. Zasady mówiące o tym, jak możesz jej używać — co jest dozwolone, a co nie — znajdziesz w Zasadach korzystania. ## Limity użycia › Limity użycia i zasady uczciwego korzystania z eustella eustella rozlicza użycie jako jeden, wspólny limit tygodniowy. Czaty, generowanie obrazów, wyszukiwanie w sieci, odczytywanie plików i uruchomienia agentów czerpią wszystkie z tej samej puli, dzięki czemu możesz wykorzystać swój limit tak, jak Ci pasuje, zamiast być ograniczonym osobnymi przydziałami dla każdej funkcji. Orientacyjnie, jeden plan Free wystarcza na około 200 czatów miesięcznie.* Te wartości są orientacyjnymi odpowiednikami tego samego limitu, a nie niezależnymi przydziałami — korzystanie z jednego pomniejsza drugi, ponieważ dzielą jedną pulę. Plany płatne dają Ci ten sam limit, tylko zwielokrotniony — więc możesz zrobić więcej jednego lub drugiego, w dowolnych proporcjach: PlanLimit tygodniowyOrientacyjnie miesięcznie (dowolne proporcje)Free1×~200 czatów / 20 obrazówComet~4×~800 czatów / 80 obrazówStar~12×~2 400 czatów / 240 obrazówCosmos~60×~12 000 czatów / 1 200 obrazów Zobacz Cennik, aby sprawdzić, co obejmuje każdy plan. Chcesz wycisnąć więcej z każdego czatu? Dowiedz się, jak najlepiej wykorzystać swój limit, w naszym przewodniku po dobrych praktykach. * Uwaga na temat tych liczb. Podane wyżej liczby czatów i obrazów to zgrubne przybliżenie — a nie rzeczywista liczba możliwych czatów, która bardzo się różni w zależności od zastosowania. eustella obliczyła je z myślą o naszych dobrych praktykach, bez żadnych tak zwanych „tool calls" (wyszukiwań w sieci, tworzenia dokumentów, zapisów w pamięci itp.) oraz przy podstawowym ustawieniu Soul w eustella. Rzeczywiste korzystanie obejmujące wywołania narzędzi, dłuższe rozmowy lub bogatszą Soul zużywa Twój limit szybciej — traktuj więc te liczby jako zgrubną orientację, a nie obietnicę. ## Limity użycia › Limity użycia i zasady uczciwego korzystania z eustella eustella rozlicza Twoje użycie za pomocą następujących elementów: Limit tygodniowy — Twój główny limit. To limit opisany powyżej. Odnawia się co tydzień.Limit kroczący 5-godzinny — krótkoterminowe zabezpieczenie (Star i Cosmos). W planach Star i Cosmos obok limitu tygodniowego działa krótszy limit kroczący, mierzony w oparciu o ostatnie 5 godzin, dzięki czemu jeden intensywny zryw nie wyczerpie go za jednym posiedzeniem. Pozostałe plany nie mają tego limitu.Doładowanie — opcjonalna pula na nadwyżki. Jeśli dodasz środki, są one wykorzystywane dopiero po wyczerpaniu Twojego zwykłego limitu — nigdy wcześniej — i przechodzą na kolejny okres, zamiast się resetować. Po osiągnięciu limitu część funkcji zostaje wstrzymana do czasu odnowienia limitu; Twoje dane i historia pozostają nienaruszone. W aplikacji Twoje użycie jest pokazywane jako procent bieżącego limitu, a nie jako narastająca liczba jednostek — powyższe wartości są orientacyjnymi odpowiednikami tego samego limitu. Niektóre konta są nierozliczane (bez limitu); na tych kontach nie widać paska użycia. ## Limity użycia › Limity użycia i zasady uczciwego korzystania z eustella › Jak działa limit Wszystko, co robi eustella, działa w oparciu o bazowy model SI, a Twój limit jest tym, co za to płaci. Dwa rodzaje aktywności go uszczuplają. Myślenie i pisanie. Każda odpowiedź oznacza, że eustella analizuje Twoje zapytanie i redaguje odpowiedź — to podstawowy koszt każdej wiadomości na czacie. Krótkie pytanie kosztuje niewiele; długie, złożone odpowiedzi i długie rozmowy kosztują więcej, po prostu dlatego, że model ma więcej do przeczytania i wytworzenia. Wywołania narzędzi. Aby działać, a nie tylko rozmawiać, eustella korzysta z narzędzi — wyszukiwania w sieci, odczytania pliku lub wspomnienia, spojrzenia na obraz, wygenerowania obrazu i innych. Zależnie od narzędzia wywołanie kosztuje albo mniej więcej tyle, ile potrzeba, aby odczytać zwróconą informację (ten tekst trafia na czat jak każde inne dane wejściowe), albo więcej, gdy samo narzędzie jest kosztowne w działaniu. W szczegółach: Odczytanie czegoś kosztuje mniej więcej tyle, co wklejenie tego. Gdy eustella odczytuje wspomnienie, otwiera plik lub ogląda przesłany przez Ciebie obraz, ta treść staje się częścią tego, co model musi przyswoić — kosztuje więc mniej więcej tyle, ile kosztowałoby wklejenie tego tekstu lub dokumentu na czat przez Ciebie. Krótka notatka jest tania; długi raport kosztuje więcej. Nic ukrytego — po prostu długość tego, co jest czytane.Wyszukiwanie w sieci kosztuje z dwóch powodów. Po pierwsze, europejski dostawca, z którego indeksu wyszukiwania korzystamy, pobiera od nas niewielką opłatę za każde wyszukiwanie. Po drugie, odczytane wyniki liczą się jak zwykłe dane wejściowe — dokładnie tak, jakby ten tekst pojawił się na Twoim czacie. Jedno szybkie sprawdzenie jest niewielkie; sesja pełna wyszukiwań sumuje się.Generowanie obrazów kosztuje więcej. Utworzenie obrazu oznacza wywołanie większego, droższego modelu generowania obrazów, więc wygenerowany obraz zużywa zauważalnie więcej Twojego limitu niż odpowiedź tekstowa. Warto o tym wiedzieć, zanim wygenerujesz tuzin wariantów.Agenci kosztują tyle, ile kosztowałoby wykonanie tej samej pracy ręcznie. Agent czerpie z tego samego limitu tygodniowego co Twoje czaty: każde wyszukiwanie, które wykona, każdy dokument, który przeczyta, każdy obraz, który utworzy, oraz raport, który napisze, liczą się tak samo jak w zwykłym czacie. Jedyne, na co trzeba uważać, to fakt, że harmonogram się powtarza — codzienny agent badawczy po cichu wydaje każdego dnia równowartość pełnej sesji badawczej, niezależnie od tego, czy przeczytasz wynik. Planuj więc rozważnie: wybierz częstotliwość, której naprawdę użyjesz, a nie najwyższą dostępną. Jeśli agent osiągnie Twój limit, jego uruchomienie zostaje po prostu wstrzymane i ponowione w kolejnym okresie, a Twój harmonogram pozostaje nienaruszony. Zabezpieczenie 5-godzinne w planach Star i Cosmos (powyżej) nadal obowiązuje. Dla porównania, zwykła wiadomość na czacie to mniej więcej najtańsza rzecz, jaką możesz zrobić — kolejny powód, dla którego krótkie, skupione czaty pozwalają Twojemu limitowi starczyć na dłużej. ## Limity użycia › Limity użycia i zasady uczciwego korzystania z eustella Jeśli wykupujesz plan płatny, przede wszystkim: dziękujemy za wsparcie. Subskrypcja daje Ci większy limit użycia — większą wspólną pulę do wykorzystania na czaty, wyszukiwania, obrazy i agentów, a więc więcej swobody, by robić więcej z eustella. Czego większy limit nie daje, to idealnych, bezbłędnych odpowiedzi. Jak każda dzisiejsza SI, eustella wciąż może się mylić lub halucynować w dowolnym planie — większy limit daje Ci więcej swobody, by pracować, wyszukiwać i uruchamiać agentów, ale nie jest gwarancją poprawności. Aby dowiedzieć się, jak uzyskiwać najlepsze i najbardziej wiarygodne wyniki z eustella, zajrzyj do naszego przewodnika: Jak korzystać z eustella — dobre praktyki. ## Limity użycia › Limity użycia i zasady uczciwego korzystania z eustella W przypadku klientów biznesowych i korzystających z API limity użycia, limity szybkości oraz przydziały określa Twój plan lub indywidualna umowa pisemna, zgodnie z Warunkami dla firm. Limity szybkości API są udokumentowane osobno dla deweloperów. ## Limity użycia › Limity użycia i zasady uczciwego korzystania z eustella eustella może z czasem korygować te limity w miarę rozwoju modeli, zmian kosztów i wzorców użycia. ## Limity użycia › Limity użycia i zasady uczciwego korzystania z eustella Wydajniejsze opcje modeli. eustella może wprowadzić opcje modeli o większych możliwościach. Jeśli tak się stanie, modele te zużywają więcej Twojego limitu na jedno działanie niż model standardowy, więc ten sam limit wystarczy na mniej działań przy modelu o dużej mocy. Powyższe odpowiedniki są szacowane dla modelu standardowego. ## Limity użycia › Limity użycia i zasady uczciwego korzystania z eustella Jeśli masz pytanie dotyczące swojego limitu lub uważasz, że limit został zastosowany błędnie, skontaktuj się z nami pod adresem support@eustella.com. © 2026 AI Newsrooms Technology GmbH. Wszelkie prawa zastrzeżone. --- # eustella — Limites de utilização URL: https://eustella.com/services/usage-limits/pt Topic: services Last updated: 2026-07-06 ## Limites de utilização Última atualização: 06.07.2026 Esta página explica como funcionam os limites de utilização na eustella, para que saiba o que esperar antes de subscrever. É disponibilizada por uma questão de transparência. Esta página trata de quanto pode utilizar a eustella. Para conhecer as regras sobre como a pode utilizar — o que é e não é permitido — consulte a Política de utilização. ## Limites de utilização › Limites de utilização e uso responsável da eustella A eustella contabiliza a utilização como uma dotação semanal única e partilhada. Os chats, a geração de imagens, a pesquisa na web, a leitura de ficheiros e as execuções de agentes recorrem todos ao mesmo conjunto, pelo que pode gastar a sua dotação da forma que lhe convier, em vez de ficar preso a quotas separadas por funcionalidade. A título de orientação, um plano Free é suficiente para cerca de 200 chats por mês.* Estes valores são equivalências indicativas da mesma dotação, não quotas independentes — utilizar uma consome a outra, porque partilham um único conjunto. Os planos pagos dão-lhe a mesma dotação, multiplicada — para que possa fazer mais de qualquer uma delas, na combinação que quiser: PlanoDotação semanalAproximadamente por mês (qualquer combinação)Free1×~200 chats / 20 imagensComet~4×~800 chats / 80 imagensStar~12×~2.400 chats / 240 imagensCosmos~60×~12.000 chats / 1.200 imagens Consulte a página de Preços para saber o que cada plano inclui. Quer tirar mais partido de cada chat? Descubra como maximizar a sua dotação no nosso guia de boas práticas. * Uma nota sobre estes números. As contagens de chats e imagens acima são uma aproximação grosseira — não o número real de chats possíveis, que varia muito consoante o caso de utilização. A eustella calculou-as tendo em mente as nossas boas práticas, sem quaisquer das chamadas «tool calls» (pesquisas na web, criação de documentos, escritas na memória e afins) e com uma configuração de Soul básica da eustella. A utilização real que envolve chamadas de ferramentas, conversas mais longas ou uma Soul mais rica consome a sua dotação mais depressa — por isso, encare estes números como uma orientação aproximada e não como uma promessa. ## Limites de utilização › Limites de utilização e uso responsável da eustella A eustella contabiliza a sua utilização com base nos seguintes elementos: Dotação semanal — a sua dotação principal. É a dotação descrita acima. É reposta todas as semanas.Limite móvel de 5 horas — uma salvaguarda de curto prazo (Star e Cosmos). Nos planos Star e Cosmos, um limite móvel mais curto, medido ao longo das últimas 5 horas, coexiste com a sua dotação semanal, para que um pico intenso não a consuma toda de uma só vez. Os restantes planos não têm este limite.Crédito adicional — um conjunto opcional para exceder o limite. Se adicionar crédito, este só é utilizado depois de esgotada a sua dotação regular — nunca antes — e transita para o período seguinte em vez de ser reposto. Se atingir o seu limite, algumas funcionalidades ficam em pausa até a sua dotação ser reposta; os seus dados e histórico não são afetados. Na aplicação, a sua utilização é apresentada como uma percentagem da sua dotação atual, e não como uma contagem contínua de unidades — os valores acima são equivalências indicativas dessa mesma dotação. Algumas contas são ilimitadas (sem limite); essas contas não veem qualquer barra de utilização. ## Limites de utilização › Limites de utilização e uso responsável da eustella › Como funciona a dotação Tudo o que a eustella faz assenta no modelo de IA subjacente, e é isso que a sua dotação paga. Dois tipos de atividade consomem-na. Pensar e escrever. Cada resposta implica que a eustella raciocine sobre o seu pedido e redija uma resposta — este é o custo de base de qualquer mensagem de chat. Uma pergunta curta custa pouco; respostas longas e elaboradas e conversas prolongadas custam mais, simplesmente porque há mais para o modelo ler e produzir. Chamadas de ferramentas. Para agir e não apenas conversar, a eustella recorre a ferramentas — pesquisar na web, ler um ficheiro ou uma memória, observar uma imagem, gerar uma imagem e muito mais. Consoante a ferramenta, uma chamada custa aproximadamente o que é preciso para ler a informação que devolve (esse texto entra no chat como qualquer outra entrada), ou custa mais quando a própria ferramenta é dispendiosa de executar. Em detalhe: Ler algo custa aproximadamente o mesmo que colá-lo. Quando a eustella lê uma memória, abre um ficheiro ou observa uma imagem que envia, esse conteúdo passa a fazer parte daquilo que o modelo tem de assimilar — por isso custa aproximadamente o mesmo que se tivesse copiado e colado esse texto ou documento no chat. Uma nota breve é barata; um relatório longo custa mais. Nada de oculto — apenas a extensão daquilo que é lido.A pesquisa na web custa por duas vias. Primeiro, o fornecedor europeu cujo índice de pesquisa utilizamos cobra-nos uma pequena quantia por cada pesquisa. Segundo, os resultados relidos contam como entrada normal — exatamente como se esse texto tivesse chegado ao seu chat. Uma consulta rápida é mínima; uma sessão cheia de pesquisas acumula-se.A geração de imagens custa mais. Criar uma imagem implica recorrer a um modelo de geração de imagens maior e mais dispendioso, pelo que uma imagem gerada consome bastante mais da sua dotação do que uma resposta escrita. Convém saber antes de produzir uma dúzia de variações.Os agentes custam o que custaria fazer o mesmo trabalho à mão. Um agente recorre à mesma dotação semanal que os seus chats: cada pesquisa que faz, cada documento que lê, cada imagem que cria e o relatório que escreve contam todos, tal como num chat normal. O único ponto a ter em atenção é que um agendamento repete-se — um agente de pesquisa diário gasta discretamente o equivalente a uma sessão de pesquisa completa todos os dias, quer leia o resultado quer não. Por isso, agende com critério: escolha a frequência que irá realmente usar, não a mais elevada disponível. Se um agente atingir o seu limite, a sua execução é simplesmente pausada e repetida na janela seguinte, e o seu agendamento permanece intacto. A salvaguarda de 5 horas nos planos Star e Cosmos (acima) continua a aplicar-se. Em contrapartida, uma simples mensagem de chat é praticamente a coisa mais barata que pode fazer — mais uma razão para que chats curtos e focados façam a sua dotação render mais. ## Limites de utilização › Limites de utilização e uso responsável da eustella Se subscrever um plano pago, antes de mais: obrigado por nos apoiar. Uma subscrição dá-lhe mais dotação de utilização — um conjunto partilhado maior para gastar em chats, pesquisas, imagens e agentes, e portanto mais margem para fazer mais com a eustella. Aquilo que mais dotação não lhe dá são respostas perfeitas e sem erros. Como qualquer IA atual, a eustella pode ainda cometer erros ou alucinar em qualquer plano — uma dotação maior dá-lhe mais margem para trabalhar, pesquisar e executar agentes, mas não é uma garantia de exatidão. Para saber como obter os melhores e mais fiáveis resultados da eustella, consulte o nosso guia: Como usar a eustella — boas práticas. ## Limites de utilização › Limites de utilização e uso responsável da eustella Para clientes empresariais e de API, os limites de utilização, os limites de débito e as quotas são definidos no seu plano ou num acordo escrito individual, conforme descrito nas Condições para empresas. Os limites de débito da API estão documentados separadamente para programadores. ## Limites de utilização › Limites de utilização e uso responsável da eustella A eustella pode ajustar estas dotações ao longo do tempo, à medida que os modelos, os custos e os padrões de utilização evoluem. ## Limites de utilização › Limites de utilização e uso responsável da eustella Opções de modelos mais potentes. A eustella poderá introduzir opções de modelos com maior capacidade. Nesse caso, esses modelos consomem mais da sua dotação por ação do que o modelo padrão, pelo que a mesma dotação chega para menos ações com modelos exigentes. As equivalências acima são estimadas com base no modelo padrão. ## Limites de utilização › Limites de utilização e uso responsável da eustella Se tiver alguma questão sobre a sua dotação ou considerar que um limite foi aplicado por engano, contacte-nos através de support@eustella.com. © 2026 AI Newsrooms Technology GmbH. Todos os direitos reservados. --- # eustella — Usage Policy URL: https://eustella.com/services/usage-policy Topic: usage-policy ## Usage Policy Last updated: 23.06.2026 These Usage Policies explain how you may - and may not - use eustella and its Services. They form part of the eustella Terms of Use. We build for trust. eustella is a European AI platform built on privacy, transparency, and safety. These Usage Policies reflect those values. They set clear expectations so that everyone can use eustella safely and responsibly. Responsible use is shared. We assume good faith from our users. These policies set a reasonable bar for acceptable use, but they are no substitute for your own legal obligations, professional duties, or ethical judgment. For guidance on the capabilities, limitations, and risks of AI - including hallucinations, over-reliance, and emotional attachment - please read our Safety Guidelines. We enforce fairly. We monitor and enforce these policies with privacy safeguards in place and clear review processes. Breaking or circumventing our rules may result in restricted access, suspension, or termination of your account. If you believe we made a mistake, you can appeal by contacting support@eustella.com, and we will work to make things right. We update as we learn. People use AI in new ways every day. We update these policies to keep pace - to avoid being overly restrictive and to better protect our users. We reserve the right to restrict access where we reasonably believe it necessary to protect our Services, our users, or anyone else. Your use of eustella must follow these Usage Policies: ## Usage Policy › eustella Usage Policies Everyone has a right to safety and security. You may not use eustella for: threats, intimidation, harassment, stalking, or defamationpromotion or facilitation of suicide, self-harm, or disordered eatingsexual violence or non-consensual intimate contentterrorism or violence, including hate-motivated violenceweapons development, procurement, or use, including conventional weapons or chemical, biological, radiological, nuclear, or explosive (CBRNE) materialsillicit activities, goods, or servicesdestruction, compromise, or breach of another person's or organisation's systems, networks, or property, including any form of malicious or abusive cyber activityreal-money gamblingprovision of tailored advice that requires a professional licence - such as legal, medical, financial, or tax advice - without appropriate involvement by a licensed professionalunsolicited security or safety testing of eustella or any connected servicecircumventing eustella's safety measures, content filters, or usage restrictions ## Usage Policy › eustella Usage Policies People are entitled to privacy. eustella is built on this principle. You may not use eustella to compromise the privacy of others, including to collect, aggregate, monitor, profile, or distribute any person's private or sensitive information without their authorisation. In particular, you may never use eustella for: building or contributing to facial recognition databases without the data subject's explicit consentreal-time remote biometric identification of persons in publicly accessible spacesusing a person's likeness - including photorealistic images or voice - without their consent, in ways that could create confusion about authenticityevaluating or classifying individuals based on their social behaviour, personal traits, or biometric data, including social scoring, profiling, or inferring sensitive attributes such as political opinions, religious beliefs, sexual orientation, or health statusinferring a person's emotions in workplaces or educational settings, except where strictly necessary for medical or safety reasonsassessing or predicting the risk that a person will commit a criminal offence based solely on personal traits or profiling These prohibitions align with the practices prohibited under the EU Artificial Intelligence Act (Regulation (EU) 2024/1689). ## Usage Policy › eustella Usage Policies Children and young people deserve special protection. eustella is designed to prevent harm and support their well-being, and must never be used to exploit, endanger, or sexualise anyone under 18 years old. We report suspected child sexual abuse material (CSAM) and child endangerment to the competent national authorities and, where applicable, through the INHOPE network. You may not use eustella for: child sexual abuse material (CSAM), whether or not any portion is AI-generatedgrooming of minorsexposing minors to age-inappropriate content, including graphic self-harm, sexual, or violent materialpromoting unhealthy dieting or exercise behaviour to minorsshaming or stigmatising the body type or appearance of minorsdangerous challenges directed at minorssexual or violent roleplay involving characters under 18facilitating minors' access to age-restricted goods, services, or activities ## Usage Policy › eustella Usage Policies People should be able to make informed decisions about their lives and communities. You may not use eustella to manipulate or deceive people, to interfere with the exercise of their fundamental rights, to exploit vulnerabilities, or to obstruct access to education or critical services. In particular, you may not use eustella for: academic dishonestydeceit, fraud, scams, spam, or impersonationpolitical campaigning, lobbying, foreign or domestic election interference, or voter demobilisationautomation of high-stakes decisions in sensitive areas without meaningful human oversight, including in: critical infrastructureeducationhousingemploymentfinancial services and creditinsurancelegal proceedingshealthcareessential government servicesproduct safetymigration and asylumlaw enforcement ## Usage Policy › eustella Usage Policies eustella can take actions on your behalf - sending messages, modifying files, interacting with connected services, and in the future performing transactions in the real world. This capability comes with responsibility. You may not use eustella's agentic features to: perform actions you are not authorised to perform, or that exceed the permissions you have been granted by a third partycircumvent access controls, authentication mechanisms, or usage limits of any connected third-party serviceinitiate transactions, purchases, or financial commitments without your own explicit authorisation or in a manner designed to obscure what is being committed tosend communications (emails, messages, calendar invitations, or any other form of outreach) that impersonate another person, that are unsolicited bulk messages, or that are intended to harass, deceive, or defraudautomate interactions with third-party services in a way that violates those services' terms of useuse connected accounts or services to access, extract, or exfiltrate another person's data without their consent You are responsible for all actions eustella takes on your behalf. Review the agent's proposed actions before confirming them, particularly for actions that are irreversible, financially consequential, or that affect other people. For more on how agentic features work and how to use them safely, see our Safety Guidelines. ## Usage Policy › eustella Usage Policies eustella is built in neutral Austria, independent by principle. eustella may not be used for: military applications, including targeting, intelligence analysis, command and control, or autonomous weapons systemsmass surveillance, bulk data collection for monitoring populations, or tracking individuals on behalf of any state or non-state actornational security or intelligence purposes without our prior written review and approvallaw enforcement profiling, predictive policing, or criminal risk scoring ## Usage Policy › eustella Usage Policies eustella generates AI content and can act in the real world on your behalf. You are responsible for ensuring that your use of Output does not infringe the rights of others. In particular: do not represent AI-generated Output as human-created when disclosure is required by law, by professional standards, or when the distinction materially affects the recipientdo not use eustella to generate content that infringes the copyrights, trademarks, or other intellectual property rights of othersdo not use Output to develop AI models that compete with eustella ## Usage Policy › eustella Usage Policies These Usage Policies set the rules for how you may use eustella. They are separate from your plan's usage allowance — how much you can use eustella in a given period. For details on weekly allowances, top-up credit, and how limits differ by plan, see Usage Limits & Fair Use. Circumventing or artificially inflating usage limits is not permitted (see also the Terms of Use). ## Usage Policy › eustella Usage Policies If you encounter content or behaviour on eustella that violates these Usage Policies, please report it to security@eustella.com. We take all reports seriously and will respond promptly. © 2026 AI Newsrooms Technology GmbH. All rights reserved. --- # eustella — Every eustella user gets up to 8× more usage — at the same price URL: https://eustella.com/blog/every-eustella-user-gets-up-to-8-more-usage-at-the-same-price Topic: ai Last updated: 2026-08-04T08:47:00Z ## Every eustella User Gets up to 8× More Usage — at The Same Price Every day, we work on eustella and improve our services for Europeans who want a more useful, capable and trustworthy AI agent in their everyday lives. That work happens across the entire product: from the models and infrastructure behind eustella to the experience of using it for writing, planning, organising, researching and getting things done. Our goal is simple: make eustella more helpful, more accessible and more dependable with every improvement we make. Today, we are proud to share an important milestone. Since launch, we have increased the number of text tokens available to every eustella user by approximately eight times. ## Same prices, 8× the tokens since launch Whether you are on the Free tier or use Comet, Star or Cosmos, your tier now includes around eight times the text token allowance that was available at launch. And the prices of all our tiers remain unchanged. This increase reflects several improvements we have made over recent weeks. Two weeks ago, we doubled the available text token allowance. With our latest improvements, we have increased it again. Together, these changes mean that every tier now offers approximately 8× the number of text tokens available at launch. We are deliberately making this improvement available across every tier. More useful AI should not be reserved for a small group of users. It should help more people bring eustella into their daily lives, whether they are just getting started or already rely on it every day. You can find the current details of our tiers and prices on our [pricing page](https://eustella.com/pricing). ## More tokens through software and hardware optimisations This progress is made possible by continuous improvements to our software and hardware. We have optimised how eustella processes requests, uses resources and makes use of the underlying infrastructure. That efficiency allows us to provide substantially more text tokens within each tier. Tokens are the units that make up the input you send to eustella and the responses it generates. More available tokens mean more conversations, longer tasks and more room to work through complex projects across multiple steps. Put simply: better software and hardware allow us to make eustella more generous, without increasing the price of any tier. ## A note on what the 8× increase includes The approximately 8× increase applies to input and output tokens for text generations. It does not apply to image generation or web searches, which have their own usage limits and resource requirements. We want to be transparent about how usage is measured, so you can understand exactly what this improvement means for your experience with eustella. ## What this means for you With significantly more text usage available, you can integrate eustella more naturally into your everyday life. You can develop ideas and texts in greater depth, organise information, prepare decisions, and plan tasks, projects and routines. eustella helps you understand complex topics, complete recurring tasks more quickly and have a thoughtful sparring partner at your side for both personal and professional questions. The Free tier also receives substantially more available usage. We want eustella to become meaningfully more useful for as many people as possible – not only after they choose a paid plan. ## European infrastructure remains our core More usage does not mean compromising on our principles. eustella continues to rely on open-weight models and European infrastructure. To achieve this, we work with European infrastructure partners such as IONOS, Scaleway and Verda. Our approach remains clear: powerful AI with a strong focus on European infrastructure, digital sovereignty and responsible data practices. We believe that useful AI and a conscious technological foundation belong together. That is why we continue to build eustella on a stack that reflects this commitment. ## This is only the beginning Reaching approximately 8× the text token allowance available at launch is an important milestone, but it is not the end of the journey. We will continue to improve eustella every day, making it more useful, accessible and capable for people across Europe. Thank you for being part of eustella. We are excited to see what you do with the additional room to think, create and get things done. --- # eustella — The Triumph of Open Weight AI – and How eustella Builds on It URL: https://eustella.com/blog/the-triumph-of-open-weight-ai-and-how-eustella-builds-on-it Topic: open weight Last updated: 2026-07-20T13:57:00Z ## The Triumph of Open Weight AI – and How eustella Builds on It Not long ago, one thing seemed settled: the world's best AI models were proprietary, locked behind the APIs of OpenAI, Anthropic, and Google – and anyone who wanted frontier intelligence paid frontier prices. In the summer of 2026, that certainty is history. Within a few weeks, two models from China have shattered it. On June 16, Z.ai (formerly Zhipu AI) released GLM-5.2, a mixture-of-experts model with around 750 billion parameters under an MIT license. It scores 51 on the Artificial Analysis Intelligence Index – the best result any open-weight model has ever achieved, just a few points behind Claude Opus 4.8 (56) and GPT-5.5 (53). On GDPval, a benchmark built around real-world work tasks, GLM-5.2 even matched a GPT flagship for the first time. And it does all this at roughly one-sixth of the API cost. ## Kimi K3 – and what comes next Exactly one month later, Moonshot AI raised the stakes with Kimi K3: 2.8 trillion parameters, making it the largest open-weight model ever announced, with a context window of one million tokens. Independent rankings put K3 on par with Claude Opus 4.8 and GPT-5.5 – only the very latest closed flagships, Fable 5 and GPT 5.6, still sit ahead. In the Frontend Code Arena, where developers compare outputs in blind tests, K3 even claimed the top spot, beating every US model. And then there's DeepSeek, the startup that already sent one shockwave through the Western world. It has raised billions in fresh capital to build its next open-weight model, expected to launch in the coming weeks or months. What it will deliver is anyone's guess. The numbers tell a clear story. The gap between closed US frontier models and open models from China has shrunk from years to – depending on the estimate – a few months. What was exclusive yesterday can be downloaded today, used commercially, and run on your own hardware. For companies that want to break free of US cloud APIs, for researchers, for Europe's sovereignty ambitions, this is a historic opportunity. For the first time, frontier-level intelligence is a freely available good. ## The flip side of openness But an honest look at this development has to include the other half of the story. Open weights mean anyone can download them – and anyone can modify them. Safety mechanisms that providers train into their models can be stripped out through fine-tuning. A model that codes at Opus level will also write malware at Opus level if you let it. Phishing campaigns in flawless German, automated vulnerability hunting, social engineering at industrial scale – the tools for all of this no longer sit behind an API with abuse detection. They sit on computers around the world. There's more: once weights are published, they cannot be recalled. No patch, no update, no kill switch. And models from China raise questions of their own – about training data, about built-in biases, about benchmark claims that independent auditors have so far only partially verified. The answer, however, cannot be to demonize open weights or pretend they don't exist. The genie is out of the bottle, and it won't go back in. The real question is: who makes this technology usable – and on what terms? ## Between open access and an open flank This is exactly where European providers come in. What's needed is a layer between raw model weights and the people who use them: an infrastructure that captures the cost advantages and independence of open models without passing their risks along unfiltered. Hosted in Europe, under European law, with abuse detection, guardrails, and clear governance – at prices not pegged to the margins of US hyperscalers. That is the gap where startups like eustella operate. Making open-weight models accessible in a sensible, affordable, and controlled way means the frontier intelligence that GLM-5.2 and Kimi K3 set free actually reaches businesses, public administrations, and private users – as a productivity tool, not a hacking kit for anyone who wants one. Openness and responsibility are not a contradiction. But they don't happen on their own. Someone has to build them. The triumph of open models is real, and it is good for Europe – if Europe shapes it instead of just watching from the sidelines. --- # eustella — "Hey eustella!" — Why We Welcome the EU Commission's Android Decision URL: https://eustella.com/blog/hey-eustella-why-we-welcome-the-eu-commission-s-android-decision Topic: europe Last updated: 2026-07-16T12:41:21.176304Z ## "Hey eustella!" — Why We Welcome the EU Commission's Android Decision The EU Commission has issued two legally binding specification decisions against Google under the Digital Markets Act (DMA). One of them matters enormously to us — and to every European AI company: Google must open Android to third-party AI assistants and grant them equal access to core functions of the operating system. In practice, that means users will be able to activate the AI assistant of their choice via voice command — just like "Hey Google" today. Third-party assistants will be able to perform actions in apps on the user's behalf: booking a taxi, suggesting replies in chat apps, answering questions about places you've visited. Exactly the kind of deep, agentic integration that has so far been reserved for Google's own AI services like Gemini. ## We welcome this decision — and we are ready Let's be clear: this is exactly the level playing field that European AI companies have been asking for, and that hundreds of millions of EU users deserve. Around 60 percent of smartphone users in the EU are on Android. Until now, whoever built the operating system also decided which AI assistant gets the privileged seat. That is changing. At eustella, we would gladly integrate deeply into Android for those hundreds of millions of EU users. Voice activation, agentic actions across apps, the full assistant experience — this is precisely what we are building a European sovereign AI agent platform for. Agents are only as useful as the places they can act in, and there is no bigger place in Europe than Android. For us, this is also about more than distribution. Digital sovereignty is often discussed in terms of where models are trained and where data is stored. But it is just as much about whether European alternatives get a fair shot at reaching users on the devices they already own. This decision is a step in that direction. ## The open questions: implementation, costs, hurdles At the same time, we are realistic. A legally binding decision is not the same as a working API. We are genuinely curious to see what the technical implementation will look like in practice: - **How accessible will the interfaces be?** Will the integration points be well-documented, stable, and open to startups — or buried in certification processes that only well-funded players can navigate? - **What will it cost?** Deep OS integration can come with fees, compliance requirements, and engineering overhead. For EU startups like us, the difference between a real opportunity and a theoretical one often comes down to exactly these practical hurdles. - **How level will the playing field really be?** Equal access on paper must translate into equal performance, equal latency, and equal capabilities on the device. The Commission has set the right direction. Whether the decision delivers real contestability will depend on the execution — and on how Google implements the measures by the July 2027 deadline. ## We'll be first in line The changes to Android are set to reach users from July 2027. Google must start sharing search data with eligible providers even earlier, from January 2027. We will be watching the implementation closely, we will engage with the process where we can — and when the door opens, we intend to walk through it. Hey eustella? We like the sound of that. --- # eustella — How to Guide: eustella best practices URL: https://eustella.com/blog/how-to-guide-eustella-best-practices Topic: eustella Last updated: 2026-07-11T06:57:00Z ## A practical guide Hi, and welcome. This is a living guide to getting the best out of **eustella**, your European AI agent. We keep it plain and honest — no marketing fog, no black boxes. It is updated regularly with new tips, features, and best practices, so it is worth checking back. A few useful links before we start: - The models behind eustella → [eustella.com/ai-models](https://eustella.com/ai-models) - Our approach to AI safety → [eustella.com/ai-models/safety](https://eustella.com/ai-models/safety) - What sovereignty means to us → [eustella.com/sovereignty](https://eustella.com/sovereignty) eustella is powerful. But like every AI agent today, she works best when you understand *how* she thinks. The habits below take five minutes to learn and will noticeably sharpen your results. ## Contents - [Hallucinations](#hallucinations) - [Keep chats focused](#keep-chats-focused) - [Use Memory](#use-memory) - [The Soul](#the-soul) - [Use Projects](#use-projects) - [Tools](#tools) - [Agents](#agents) - [Usage Allowance & Limits](#usage-allowance--limits) ## Hallucinations *What they are, why they happen, and why nobody has "fixed" them.* Start here, because this one shapes everything else. Sometimes an AI (and in our case, a LLM) states something that sounds completely confident — a street address, a bus line, a date, a quote — and it turns out to be wrong. This is called a **hallucination**. We would rather be honest than reassuring: it is not a bug we forgot to fix. It is a fundamental, still-unsolved property of how today's language models work. A model does not "look things up" in a table of facts the way a librarian would. At its core it predicts, one small piece at a time, the *most plausible* continuation of the text so far, learned from enormous amounts of data. Most of the time the most plausible answer is also the correct one — which is exactly why it feels so capable. But plausible and true are not the same thing. When the model is unsure, it does not go quiet. It produces the most likely-*sounding* answer, and delivers it with the same confidence as a fact it is certain about. eustella works hard to narrow this gap. She searches the live web. She checks sources. She can remember facts you save in Memory. These steps genuinely lower the error rate — but they do not remove it. No provider on earth has eliminated hallucinations, not even the largest US labs with the largest models. Anyone who tells you otherwise is overselling. **What this means for you:** treat eustella as a brilliant, fast, tireless assistant whose factual claims are worth a quick check when the stakes are high — exact addresses, legal or medical details, numbers you will act on. Ask her to search and cite her source. Everything that follows is really about one thing: habits that keep the hallucination rate low. ## Keep chats focused *Short, clean conversations. Start new ones often.* This is the single easiest way to get better answers. Every model has a **context window** — the amount of text (measured in "tokens", roughly word-pieces) it can hold in mind at once. Everything in your conversation fills that window: your messages, eustella's replies, the documents and search results she pulls in. The fuller it gets, the harder the model must work to track it all, and the easier it is to lose a detail, mix two things up, or drift from your original instructions. In practice: the longer a single chat runs, the higher the chance of a hallucination. When a conversation grows very long, eustella automatically summarises the earlier part to make room. That is necessary — but a summary is never as faithful as the original. Small errors can slip in, and then be repeated. We will be straight about why limits exist at all. eustella runs on **open-weight models** and is a **European provider**. We do not have a hyperscaler's effectively infinite compute, and we do not have billions of dollars to spend on unlimited token budgets — and honestly, we think burning vast amounts of energy for marginal gains is the wrong trade for Europe. So we run lean and set sensible limits. The upside is real and on your side: shorter, focused chats are not only cheaper to run — they genuinely produce more accurate answers. **Best practice:** finished a task, or switching topic? Start a **new chat**. Treat each chat as one clean workspace for one job. And do not worry about losing what matters between chats — that is exactly what Memory is for. Learn how to use it below → ## Use Memory *Carry the important things with you, chat to chat.* If you start fresh chats often, how does eustella remember your context? Through **Memory**. Just ask, in plain language: > "Please remember that I run a small bookshop in Graz and prefer supplier emails written formally." eustella saves that as a durable memory and draws on it in future chats, so you never repeat yourself. You can ask what she currently remembers, tell her to update or delete something, and review her memory whenever you like - either via the Chat or via the memory window in the application. Another way to make eustella remember useful information is after a round of trial & error. For example: > "Now remember XYZ / everything that we just discussed / the exact way on how I want XYZ" There are many ways to organise it, e.g.: - **You structure it** — tell eustella exactly what to keep and how ("keep a list of my active projects"; "store my brand-tone rules separately from my personal preferences"). - **eustella structures it** — just tell her what matters and let her organise the memory system sensibly for you. Good memory habits are what let your chats stay short and sharp without ever losing the thread of your work. **A few memory habits that pay off:** - **Be specific.** "I prefer metric units and ISO dates" beats "I like things tidy." Precise memories get used precisely. - **Audit it now and then.** Ask "what do you remember about me?" to see everything at a glance, and prune anything out of date — a stale fact can mislead her the same way an old note misleads you. - **Lock in a hard-won result.** After you finally get a tricky task exactly right, say "remember precisely how we did this" so you never have to rebuild it from scratch. - **Save the durable, skip the throwaway.** Keep the things that will still be true next month; let one-off details stay in the chat where they belong. **Pro Tip**: Tell eustella to reference or mention important memories in her *Soul* — that way they shape how she works, not just what she knows. ## The Soul *Who eustella is, how she does things, and who she thinks you are.* If Memory is what eustella knows, the **Soul** is who eustella *is*. It is the heart of her — not a settings page, but the definition of the agent you are talking to. The Soul can hold many things and you decide. Here are three examples of what could be important to add to the Soul: - **Who she is** — her character, her values, the way she speaks and carries herself. This is where you shape eustella's personality into one that fits you. - **How she does things** — her working style and standing rules: when and how to search, how to correct herself, what to remember, the tone and format she replies in. - **Who she thinks you are** — her picture of you: your role, your context, what you care about, how you like to work. This is what lets her help fit your life rather than a generic user's. You shape all of it in plain language, and it holds across every chat. For example: - "Always verify addresses and opening hours with a live search before giving them to me. Add that to your Soul!" - "Write in your Soul: If a search result contradicts something you said earlier, trust the source and tell me you changed your answer." - "Update your Soul so that whenever I mention a client, save their details to Memory." - "I only speak Spanish. So you should always respond in Spanish. Put that in your Soul." The Soul is what turns eustella from a capable tool into *your* agent. Combined with the habits above — verify facts, keep chats focused — a well-shaped Soul is what makes eustella genuinely dependable. **Getting the most from the Soul:** - **Keep it sharp, not sprawling.** A short, well-chosen Soul works better than a long, rambling one — the same reason a crisp brief beats a wall of text. Add what genuinely matters, and prune the rest. - **When you repeat yourself, promote it.** Caught yourself giving the same instruction twice? That is the signal it belongs in the Soul, not in yet another message. - **Read it back.** Ask "what's in your Soul right now?" to check she is carrying the version of herself — and of you — that you actually want. - **Divide the labour with Memory.** Soul for *who she is and how she works*; Memory for *the facts*. Together they mean she starts every chat already knowing you. ## Use Projects *The right home for document-heavy work.* Working with a lot of files — reports, contracts, research PDFs, notes? Put them in a **Project**. A Project keeps a set of documents together so eustella can draw on all of them across your chats, without you re-uploading or re-pasting anything. It is the natural home for a research dossier, a legal matter, a book you are writing, or a business plan. Keep unrelated work in separate Projects, so each one stays focused — the same principle as keeping chats clean. **Best practices for Projects:** - **One Project, one purpose.** A Project works best when its files pull in the same direction — a single client, a single book, a single case. Mixed-topic Projects dilute her focus, just like an over-full chat does. - **Give her a way in.** Drop a short introduction or brief into the project settings describing what the Project is and what you want from it. It orients eustella before she opens a single file. - **Keep chats fresh inside a Project.** Your documents stay put no matter how many chats you start — so start new ones freely (remember: short chats, fewer hallucinations). The Project is the memory; each chat is a clean workspace. **Pro Tip:** point an **agent** at a Project (see below) and eustella will work through your documents on a schedule — for instance, a weekly digest of everything new in a case file. ## Tools *What eustella uses to get things done.* eustella does not just talk; she acts. To do that she reaches for **tools**: searching the live web, generating images, reading the files in your Projects, and more. When a task needs one, she picks the right tool, uses it, and works the result into her answer. **How do I know what she can do? Just ask her.** "What can you do?", "What tools do you have?", or simply "Can you do X?" — she will tell you what is available and how to use it. You never have to memorise a tool list: describe the outcome you want, and eustella reaches for the right tool herself. **Can I add my own tools?** Not directly yet — today the toolset is the one eustella ships with. But that is about to change (see below with the MCP Servers comming soon), and it is a feature we are very excited about. **The real power is combining tools with everything else.** Tools become far more useful when they work together with Memory, the Soul, and Projects: - **Tools + Soul** — "always search before quoting a price" turns a one-off request into a standing habit she never forgets. - **Tools + Memory** — she searches, then saves what she found, so your next chat starts from it instead of from zero. - **Tools + Projects** — she reads your own documents *and* checks them against the live web in the same breath. That interplay — act, remember, apply your standards, work on your files — is where eustella stops feeling like a chatbot and starts feeling like an agent. **Coming soon — connect your own MCP server.** We are building the ability to connect your own **MCP server** to eustella. In plain terms: this will let her securely reach *your* tools and data — your systems, your databases, the services you already use — and act on them directly, under your control. Your data stays on your terms, not ours. It is a real step toward eustella working *inside* your workflow, not just alongside it. We are actively working on it — more when it lands, so watch this space. ## Agents *Put eustella to work in the background — even while you sleep.* An **agent** is eustella working on her own, in the background, on a job you hand her. You do not have to sit and wait: she does the work and tells you when it is done. eustella comes with ready-made agents — **Deep Research** (multi-step web research written up as a proper report), **Trip Planner** (flights, stays, and activities turned into an itinerary), a **Morning Brief**, a **Watchlist** that follows topics or companies and reports what changed, and more. You can also **build your own**: give a custom agent a name, a set of instructions, and the tools it may use, and it joins your line-up. **Schedule them.** Any agent can run on a schedule — **once** at a set time, on a simple **interval** (every few hours, daily), or on a precise recurring pattern, all in your own timezone. This is how you get a research digest every Monday morning, or a market watch each evening, without lifting a finger. When a scheduled run finishes, eustella **notifies you and opens a review chat** with the results — the report and its sources are saved to your Memory, ready to read, question, or build on. Nothing is dumped at you; you open it when you want. **Best practices for agents:** - **Start with a ready-made one.** Run Deep Research or Trip Planner once, on demand, to get a feel for how agents work before you build or schedule anything. - **Give a custom agent a clear brief.** The sharper the instructions and the tighter the set of tools you grant, the more predictable and useful the result. Treat it like briefing a capable new colleague. - **Give each research its own question.** Distinct, well-scoped questions produce distinct, well-scoped reports — so ask "compare A vs B for my use case" rather than re-running the same broad prompt. - **Schedule for what you will actually read.** A weekly digest you act on beats a daily one you ignore — and it is kinder to your allowance, since an agent spends from it just as a normal chat would. Pick the cadence that matches your real routine. **And it only gets better from here.** Agents are young and improving fast — expect richer scheduling, more ready-made agents, and, once MCP connectors land, agents that can reach *your* own tools and data. What they can do today is just the beginning. ## Usage Allowance & Limits *How much you can use eustella.* Almost everything eustella does — every chat, search, generated image, file she reads, and agent run — draws from your **usage allowance**. How large that allowance is depends on the plan you are on (Free, Comet, Star, Cosmos): the more you pay, the more you get, to spend in whatever mix suits you. We meter usage for one honest reason — running capable models on European infrastructure costs real money and energy. In the app your usage shows as a simple **percentage of your current allowance**, not a running bill. Have a tip you would like to see covered, feedback on the guide, or a question it does not answer yet? Write to us at **[support@eustella.com](mailto:support@eustella.com)**. This guide grows with what you tell us. --- # eustella — Apple, Open Up! Why Eustella Must Become the Sister of Siri AI URL: https://eustella.com/blog/apple-open-up-why-eustella-must-become-the-sister-of-siri-ai Topic: eustella Last updated: 2026-07-08T13:46:27.363872Z ## Apple, Open Up! Why Eustella Must Become the Sister of Siri AI You have to give Apple credit for this: what the company presented at WWDC 2026 with Siri AI is technically impressive. After two years of delay, the once-mocked voice assistant has become a genuine AI agent. Siri AI understands personal context from messages, emails, and photos, recognizes what's happening on screen, and orchestrates workflows across app boundaries. A system orchestrator taps into the Spotlight index and the App Toolbox, while the Foundation Models — developed in cooperation with Google's Gemini — run on-device and via Private Cloud Compute. That's the future of the smartphone interface. And to be clear upfront: we're genuinely big Apple fans. Hardly any company builds hardware and software so consistently as one piece, and hardly any takes privacy as seriously. That stance deserves respect, and it's a key reason why millions of Europeans love Apple devices and trust them with their most personal data. But that's exactly where the problem begins. Because eustella, as the new European AI agent also available for iPhone, would be at a purely technical disadvantage compared to Siri AI. **The Unfair Home-Field Advantage** No other AI agent on the iPhone gets anywhere near this level of access. They all live in their app sandbox. They can't see what's happening on screen. They can't look through the iMessage history to find last week's hotel booking. They can't trigger an action in a third-party app without the user switching to that app and typing it in themselves. Siri AI can do all of that. Not because Apple's models are better (tellingly, they come from Google and are Gemini distillations), but because Apple grants itself operating-system-level interfaces that no one else gets. Specifically, these are the capabilities that make the difference between Siri AI and every other AI app on the iPhone: **Personal Context**: Siri AI searches personal messages, emails, photos, and documents, finding things like last week's hotel booking number or restaurant recommendation. An AI app sees none of that — it only knows what the user tells it. **Onscreen Awareness**: The assistant recognizes what's currently displayed on screen, answers questions about it, and directly executes matching actions. For an AI app in its sandbox, the rest of the screen is invisible — it would have to ask for screenshots. **App Orchestration**: Siri AI moves across app boundaries and completes entire workflows — from looking up an address to booking in a third-party app. An AI app can at best open a link and then leave the user to keep tapping on their own. **Visual Intelligence**: Deeply integrated into the Camera app, Siri AI analyzes the real world through the lens — including actions like splitting a bill via Apple Cash. AI apps have to take the detour of uploaded photos. **Access points**: Siri AI is everywhere — side button, "Hey Siri," Dynamic Island, Spotlight, system-wide context menus, Apple Watch, CarPlay, AirPods. An AI app is one icon among many that the user has to actively open. Siri AI's competitive edge, then, isn't an innovation achievement. It's a platform privilege. In economics, this is called self-preferencing: the gatekeeper favors its own service on its own infrastructure. This is precisely what the Digital Markets Act was written against. Anyone classified as a gatekeeper may not equip their own services with exclusive system access that's denied to competitors. Interoperability isn't a request — it's an obligation. **No EU Launch for Siri AI** Apple has announced that it will not roll out Siri AI on iPhone, iPad, and Apple Watch in the EU for now, accompanying this with the familiar narrative: regulation endangers users' privacy and security, and the company is "working hard on a way forward." Translated, that means: Apple would have to make the interfaces that make Siri AI so powerful — Personal Context, Onscreen Awareness, App Orchestration — available to third parties too. And Apple apparently doesn't want that. The company would rather forgo 450 million potential users than share the platform privilege. One can lament this as a competitive disadvantage for the location, and yes: for European iPhone users, it's annoying in the short term that the most exciting Apple feature in years is passing them by. But the alternative would be worse. A Siri AI that sits exclusively and without competition at the deepest level of the operating system would end the AI assistant market on the iPhone before it even began. Once someone uses the assistant that knows all their personal data and can control all their apps, they won't switch away. The lock-in would be extremely strong. **What Needs to Happen Now — and Where eustella Comes In** The second step is still missing, though. It's not enough for Siri AI to pause in the EU. The DMA's goal was never to block features, but to open them up. The Commission should therefore give Apple a clear path: Siri AI may launch in Europe as soon as the underlying interfaces are available to third parties without discrimination. Personal Context, onscreen context, system orchestration: whatever Apple grants itself must also be open to others via API. Just as NFC was opened up for other payment apps alongside Apple Pay, the iPhone must also open up to other AI agents. A European assistant with access to Personal Context and app orchestration, but with data processing under European law, on European infrastructure, without a detour via Californian or Chinese servers: that would be the true sister to Siri AI. Not as an Apple replacement, but as a choice. The user decides for themselves which assistant gets to operate at the system level, just as they can already choose their browser and search engine today. That freedom of choice is precisely the logic of the DMA, thought through to its conclusion. Of course, Apple will object that deep system access for third parties is a security risk. That argument can't simply be dismissed, and that's exactly why Europe should impose high, verifiable requirements on anyone who wants to use these interfaces: audits, certification, clear liability. Private Cloud Compute shows that verifiable privacy architectures are possible. They just must not remain Apple's monopoly. Siri AI is a promising product from a company we like a great deal. It would be an even better one if it had competition on equal footing. The DMA has opened the door for that. Now Europe must walk through it, and European providers must be ready to step up. We are ready. --- # eustella — eustella Launch Event: "A Jolt Is Going Through Europe" URL: https://eustella.com/blog/eustella-launch-event-a-jolt-is-going-through-europe Topic: eustella Last updated: 2026-07-08T13:43:13.096788Z ## eustella Launch Event: "A Jolt Is Going Through Europe" It was hot, it was packed, and it was exciting: on Wednesday evening at the AI Factory in Vienna, 150 attendees — journalists, AI entrepreneurs, investors, beta testers, and representatives of various organizations — gathered for the presentation of the new AI agent platform eustella, which is fully committed to European tech sovereignty, to hear CEO Matteo Rosoli and his guests on stage (more on eustella here). "The market isn't waiting for tomorrow's perfect solution — it needs today's sovereign solution. With eustella, that's now available to everyone who no longer wants to compromise on data privacy," said Matteo Rosoli, CEO of eustella. "eustella shows what happens when AI expertise, sovereign infrastructure, and a strong ecosystem come together. That's exactly why AI Factory Austria AI:AT was built: we create access to expertise, guidance, infrastructure pathways, and partners. The fact that a startup headquartered in our coworking hub is bringing a European AI solution to market is a strong signal for the location," said Karl Kugler, Co-Lead of AI Factory Austria AI:AT. Also on the panel: Johannes Brandstetter, Vice President AI for Science at Mistral AI, along with Phillip Maasberg, Global Startup Program Lead at IONOS Cloud, and Christian Trummer, Chief Scientist & Co-founder of Bitpanda. As part of the partnership with Bitpanda, real-time cryptocurrency price data will be displayed directly within eustella. "We firmly believe that in the future, agents will interact with various services just like humans do — including in the financial sector. That's why we want to become agent-friendly — that's our goal. One step toward that is the partnership with eustella: you offer agents and operate in Austria and Europe — I see great potential for a really good collaboration there," said Trummer about the cooperation with eustella. **"The problem isn't new, but geopolitics has made it extremely acute right now"** "In Europe, we finally need to respond as Europeans — across borders, from startup funding to unified policy and defense. Otherwise we'll tear ourselves apart. For 20 years we haven't produced our own major tech companies, we're no longer leading, the German auto industry is collapsing. The problem isn't new, but geopolitics has made it extremely acute right now," said Brandstetter, who built Emmi AI in Linz and then exited to Mistral from France. "Especially with data, people react strongly because it affects everyone personally. But something is happening: a jolt is going through Europe. Now is the moment to build — and not let it fade again." The fact that eustella now offers a European challenger to ChatGPT was met with widespread approval. "When you compare Americans and Europeans, the Americans have a huge advantage: they focus on the user. And that obviously works. eustella is now stepping up and saying: we focus on the user — and make it cheaper too. If that pays off, it'll get really interesting," said business angel Hermann Futter, who is an investor in Newsrooms.ai, the company behind eustella. And further: "Right now, an enormous number of people are coming into contact with AI for the first time. Sure, a lot of people have ChatGPT — but that's our bubble, that's the students. I'm talking about the other 80% who don't have a solution yet. They're coming now too — and they're encountering a product that's affordable, that's European, and that they don't need to be afraid of." --- # eustella — eustella: European AI Agent Now Generates Slides and Works With Documents URL: https://eustella.com/blog/eustella-slides-and-documents Topic: beta Last updated: 2026-06-08T00:00:00Z ## eustella: European AI Agent Now Generates Slides and Works With Documents *Originally published by [Trending Topics](https://www.trendingtopics.eu/eustella-presentations/) on 8 June 2026.* The Vienna-based AI startup eustella has rolled out a feature update for its users. Positioned as a European "AI companion" or AI agent and currently in an open beta phase, the service is gaining primarily document-related capabilities for creating and editing office files. Among the new features is the ability to create and manage projects within the app. In addition, eustella can now read, update, and generate files in PDF, Word, Excel, and PowerPoint formats. A key focus of the update is the adoption of design templates: users can upload Word and PowerPoint files and train eustella on their respective layouts. On request, a template can be saved as a "theme." The system is then said to be able to create new slides and Word documents in the exact design of the uploaded file. The update also integrates an image gallery into the application's memory section. Beyond that, the company cites general improvements to eustella's capabilities without breaking them down in detail. ## Web version and image generation via FLUX.2 eustella had already significantly expanded its feature set with the previous update, which marked the transition into open beta alongside the launch of the web version. At its core was image generation via the FLUX.2 model from German provider Black Forest Labs, as well as the ability to analyze uploaded photos. This was accompanied by initial skills offering specialized expertise — for instance in the area of design — reworked group chats, expanded memory management, and an improved search function. The new option to adopt design templates from office files builds directly on these design skills. [eustella](/) is available as a web application as well as for [Android via the Google Play Store](https://play.google.com/store/apps/details?id=com.eustella) and — through TestFlight — for [iPhone in the App Store](https://testflight.apple.com/join/kHfpZmRJ). As the service is in open beta, the team continues to call on its testers to share feedback via a [survey](https://tally.so/forms/VLodxy) and the project's [WhatsApp community](https://chat.whatsapp.com/BWyPCzhF850DlXWYhqAvjG). --- # eustella — eustella Opens the Beta: Web App Launches with 1,000 New Spots URL: https://eustella.com/blog/eustella-web-public-beta Topic: beta Last updated: 2026-05-22T00:00:00Z ## eustella Opens the Beta: Web App Launches with 1,000 New Spots eustella opens the beta to a wider audience today. As of 22 May 2026, the eustella web app is live at [eustella.com](https://eustella.com), and **1,000 new beta spots** are now available. Limited places, first come first served. 2,500 Europeans have been using eustella through the closed beta so far — across every country in the EU, in almost every language, with a super active community shaping the product day after day. Today we are opening the door to the next wave. ## What is new **Web app.** eustella now runs in any modern browser — Chrome, Firefox, Safari, Edge — on desktop, laptop, and tablet. No install, no app store, no waiting in line. Open eustella.com, sign in, and the assistant is one click away. **1,000 new beta spots.** We are opening a fresh round of beta access alongside the web launch. Spots are limited and will be allocated in the order people sign up. **Same product, more surfaces.** The Android app on Google Play and the iOS app (in final App Store review) keep running alongside the web app. Conversations, agents, scheduled tasks, and integrations sync across every surface — start something on your phone, finish it in the browser. ## What you can do on day one The web app ships with everything beta members have been using: - **Hyperpersonalised chat** — eustella remembers your context across sessions and learns your preferences over time. - **Agents** — create custom agents for recurring workflows, research, drafting, or analysis. - **Scheduled tasks** — let eustella run jobs in the background and report back. - **Tool integrations** — connect eustella to the services and data sources you already use. - **Almost every language** — eustella chats in the language you write to it, from the major European languages to most of the world's most widely spoken ones. - **File and document analysis** — upload, summarise, extract, translate. - **Voice input** — start a conversation by speaking, right from the chat bar. Everything runs on European and open-source AI models, in EU data centres, with no training on user conversations. ## A European community eustella is in active daily use across every country in the European Union. The beta community runs conversations, builds agents, automates tasks, and pushes the product in directions we did not predict — in almost every language they write to it. That community is the reason eustella ships as fast as it does. ## What stays the same eustella is still built in Europe, for Europe, by [newsrooms.ai](https://newsrooms.ai), an independent AI company based in Vienna, Austria. No Big Tech investors. No military contracts. No training on your conversations. All data stored and processed in the EU, under GDPR and the EU AI Act. The product is still labelled beta because we are shipping fast — weekly updates, new features, and rapid iteration on user feedback. Expect rough edges; expect them to be fixed quickly. ## How to claim a spot Open [eustella.com](https://eustella.com), click **Join our Beta!**, and register with your email. The 1,000 spots are allocated in signup order, so earlier is better. Existing beta accounts work in the web app with the same login. If you miss this round, the [newsletter](https://eustella.com/newsletter) is the best way to hear about the next one. ## Thank you To the 2,500 Europeans who have been with us through the beta so far — thank you. Your conversations, your bug reports, your feature requests, your patience with the rough edges, and your willingness to use a European AI before it was the easy choice are what made today possible. The web app, the new spots, the roadmap ahead — all of it is built on what you helped us learn. See you in the chat. --- # eustella — Waitlist Closed: eustella Rolls Out Closed Beta to 2,000+ Signups URL: https://eustella.com/blog/eustella-closed-beta-download-rollout Topic: beta Last updated: 2026-04-22T00:00:00Z ## Waitlist Closed: eustella Rolls Out Closed Beta to 2,000+ Signups eustella is closing its public waitlist. More than 2,000 Europeans signed up to test the secure European alternative to ChatGPT, and the closed beta is now rolling out. As of today, 22 April 2026, new signups to the waitlist are paused. Download links are being sent in waves to everyone who registered, so existing waitlist members do not need to take any action \- the invite will reach the inbox you signed up with. ## What is live right now The eustella Android app is listed on Google Play and available to waitlist members as a closed beta. Access is rolled out progressively so we can monitor stability, infrastructure load, and user feedback before widening the pool. Waitlist members receive: - A personal installation link from Google Play - A short onboarding guide for first-time setup - A direct channel to the eustella community The iOS version for iPhone and iPad is in final review and will follow on the App Store in the coming weeks. iOS waitlist members will be notified by email as soon as it is available. ## Why we closed the waitlist now We deliberately kept the first beta cohort small. 2,000+ users is enough to stress-test eustella under real conditions \- across eight European languages, a wide range of devices, and a broad set of real use cases \- without overloading the European infrastructure we run eustella on. Closing the waitlist also lets us focus on shipping updates quickly for the current cohort rather than onboarding new users at scale before the product is ready for a public launch. ## What happens next **Next few weeks.** Closed beta users get rapid iteration \- weekly updates, bug fixes, and early access to new features. **After the beta.** eustella will open to the public with a wider release, expanded feature set, and additional languages. Everyone who missed the waitlist will be notified through our newsletter when public access opens. If you missed the waitlist, the [newsletter on eustella.com](https://eustella.com) is now the best way to stay informed. Subscribers will be the first to hear about the public release and any additional beta waves. ## What eustella is, in short eustella is a personal AI assistant built in Europe, for Europeans. It runs on European and open-source AI models, stores data exclusively in the EU, and does not train on user conversations. It is built by [newsrooms.ai](https://newsrooms.ai), an independent AI company based in Vienna, Austria. Thank you to everyone who signed up. See you in the app. --- # eustella — Why European AI Startup eustella Uses Chinese Open-Source Models URL: https://eustella.com/blog/why-eustella-uses-chinese-open-source-models Topic: open-source Last updated: 2026-04-10T00:00:00Z ## Why European AI Startup eustella Uses Chinese Open-Source Models *Originally published on [Trending Topics](https://www.trendingtopics.eu/eustella-china-open-source-ai/) on 10 April 2026.* At first glance, it sounds like a contradiction — perhaps even a PR problem. A European AI startup developing a sovereign smartphone agent chooses to build on models from China. Ten days ago, eustella announced its plan to build a European AI agent for smartphones, offering more than 100 million European AI users an alternative to ChatGPT, Claude, and others. The decision to use open-source models from DeepSeek, Alibaba's Qwen, or Moonshot looks like a strategic mistake in the current geopolitical climate. On closer inspection, it is the exact opposite. ## The question is not where a model comes from — it is who controls it The reflex is understandable. For months, geopolitical headlines have dominated the AI debate: export controls, chip embargoes, security reviews, the struggle for digital sovereignty. In this climate, integrating open models of Chinese origin into a European product seems questionable. But the real question is not where a model was built. It is who operates it. A closed model from California, accessed through an API, sends every single request outside European territory — along with user data, context information, and everything else that makes a smartphone agent useful. An open model from China, whose weights are freely available, can be hosted on European infrastructure, audited, fine-tuned, and if necessary taken offline entirely. ## Why China is aggressively pursuing open source Chinese companies are not opening their models out of altruism. It is a strategic calculation with several motives: - **Circumventing export controls.** Without access to the most powerful Nvidia chips, Chinese labs cannot compete with OpenAI, Anthropic, or Google in a pure compute arms race. Open source serves as leverage to build global relevance through reach and community rather than raw computing power. - **Free R&D at scale.** Opening a model generates free feedback, fine-tuning, benchmarks, and tooling from tens of thousands of developers worldwide. For latecomers, it is the fastest way to close the gap. - **The model is not the product.** Alibaba makes its money from cloud and commerce, Tencent from gaming and WeChat, Baidu from search. The model itself does not need to be monetised — it serves as a lead generator for cloud services or as a strategic asset. - **Commodifying the competition.** When Chinese labs distribute strong models for free, they compress the prices and margins of GPT-class APIs worldwide. DeepSeek-V3 and R1 triggered exactly this effect — including sharp stock market reactions against Nvidia and Western AI equities. - **Geopolitical influence.** Open models of Chinese origin running in products worldwide are a geopolitical statement. They give China influence over technical standards, tokenisers, and language coverage. In China's brutal AI market — the so-called "war of a hundred models" — a top benchmark result on Hugging Face doubles as a recruiting and reputation tool. For eustella, these motives are secondary. What matters is the consequence: the interest of Chinese labs in commodifying American closed-source models happens to align with Europe's interest in technological independence. You just have to seize the opportunity. eustella does not rely exclusively on Chinese AI. Google, OpenAI, Mistral, and potentially soon Meta all offer strong open-source LLMs that can be deployed depending on user needs. ## The security concerns are real — but they concern something else Open model weights are not remotely controllable. They do not phone home. They contain no backdoors that activate with the next update — an open model does not receive updates. It is a file. What does need to be examined is bias, censorship behaviour, gaps in training data, and how a model responds to sensitive topics. These are solvable problems: through fine-tuning, through evaluation, through combining multiple models, through clear guardrails in the agent layer above. And it is precisely there — in orchestration and product design — that an AI startup creates its value, not in the base model. ## The real controversy The real controversy should not be that a European startup uses Chinese models. The real controversy should be that Europe spent years talking about building its own foundation models and ended up with a handful of labs that, despite billions in announcements, have not delivered what has long been available on Hugging Face. Mistral is the honourable exception, but one swallow does not make a summer. As long as that is the case, it would be negligent to reflexively reject the best alternatives to American API dependency. eustella is therefore taking the only pragmatic path: the company takes the best open models the market offers — regardless of whether they come from China, the United States, or Europe — runs them under European control, and builds an agent whose value lies in the product, not in the base model. The truly European answer to the AI race is not choosing between Silicon Valley and Beijing. It is taking what is open from both camps and building something of your own. --- *Source: [Trending Topics — "Warum das europaische KI-Startup eustella auf chinesische Modelle setzt"](https://www.trendingtopics.eu/eustella-china-open-source-ai/), 10 April 2026.* --- # eustella — Almost Two-Thirds of Europeans Back Replacing US Tech, Poll Finds URL: https://eustella.com/blog/two-thirds-europeans-back-replacing-us-tech Topic: digital-sovereignty Last updated: 2026-04-03T00:00:00Z ## Almost Two-Thirds of Europeans Back Replacing US Tech, Poll Finds Around two-thirds of Europeans support replacing American technology services with European alternatives, according to a YouGov poll. The survey covered more than 1,000 respondents in each of five EU countries — France, Germany, Spain, Italy, and Poland — and was conducted between 6 and 16 March 2026. The poll, [reported by Tech Policy Press](https://www.techpolicy.press/almost-two-thirds-of-europeans-back-replacing-us-tech-poll-finds/) based on YouGov data, found that over 60 percent of respondents consider it a good idea to replace US services for data storage, video conferencing, and payment systems with European competitors. Specifically, 62 percent favoured switching data storage and payment services, while 59 percent backed moving away from American video conferencing providers like Zoom. Only 13 percent of participants called such a move a bad idea; another 25 percent were undecided. ## Tensions Between Washington and Brussels The poll comes against the backdrop of growing conflict between the EU and the United States over digital policy. US President Donald Trump claimed — without evidence — that Europe's online safety regulations restrict the free speech of Americans. In August 2025, he wrote on social media: "I will stand up for countries that attack our incredible American technology companies" — without naming the EU explicitly. He added: "Digital taxes, Digital Services Acts, and Digital Markets regulations are all designed to harm or discriminate against American technology." The European Commission rejected these allegations. It is conducting ongoing proceedings in the areas of online safety and competition against both US and Chinese technology companies. ## Broad Support for European Alternatives Support for switching ran across all categories surveyed. Across all five countries, 62 percent favoured replacing US data storage and payment systems; 59 percent backed switching away from US video conferencing solutions. Support was particularly strong in Italy and Germany, where 67 and 68 percent of respondents respectively said replacing servers and data storage was a good idea. France came in slightly lower at 59 percent, while Poland brought up the rear at around 50 percent. Only 13 percent of respondents overall rated a switch as a bad idea — a remarkable consensus across the EU's most populous member states. Poland stood out as the only outlier. There, support hovered around the 50 percent mark across all digital services, and 38 percent of respondents had not yet formed an opinion — significantly more than in the other four countries. This high share of undecided respondents suggests that the debate over digital sovereignty has not yet reached the same intensity in Eastern Europe as in the West. ## Germany Sceptical, France Optimistic Despite broad support in principle, many Europeans doubt the feasibility. Across all five countries, 41 percent of respondents considered a switch unrealistic, while only 40 percent considered it realistic; 19 percent offered no opinion. Germany showed the strongest pessimism: 51 percent of German respondents — the highest figure of any country — called leaving US technology unrealistic. This is notable because Germany co-hosted a digital sovereignty conference with France late last year. At the same time, only 12 percent of Germans said such a switch was a bad idea — a contradiction that highlights the gap between aspiration and scepticism. ## The Gap Between Desire and Reality The French were the most optimistic: 45 percent considered a switch realistic, while only 32 percent called it unrealistic. Yet even in France, a significant gap exists between willingness and awareness. In January 2026, Paris announced plans to replace services such as Zoom and Microsoft Teams with a French video conferencing solution for government use by 2027 — partly to regain control over critical digital infrastructure. However, 89 percent of French respondents had heard little or nothing about these plans, despite the government's active promotion. Overall, only 11 percent of French participants said they were aware of the initiative. Amid these tensions, the European Commission defended its right to regulate the digital single market. Spokesperson Paula Pinho told journalists: "It is the sovereign right of the EU and its member states to regulate economic activities on our territory that are compatible with our democratic values." French President Emmanuel Macron has been among the most vocal advocates for an independent European path in technology, including the development of homegrown artificial intelligence. The poll results show that growing frustration between transatlantic partners is increasingly reaching ordinary European citizens — even if the path from intention to implementation remains long. --- *A YouGov poll of respondents in France, Germany, Spain, Italy, and Poland shows that around 60 percent of Europeans support replacing American technology services with European alternatives. Yet only 40 percent consider the move realistic — a reflection of the growing transatlantic tensions in digital policy.* *Source: [Tech Policy Press — "Almost Two-Thirds of Europeans Back Replacing US Tech, Poll Finds"](https://www.techpolicy.press/almost-two-thirds-of-europeans-back-replacing-us-tech-poll-finds/), 26 March 2026.* --- # eustella — eustella CEO Matteo Rosoli Pitched at TRANSFORM 2026 in Berlin URL: https://eustella.com/blog/eustella-ceo-transform-2026 Topic: transform Last updated: 2026-03-24T00:00:00Z ## eustella CEO Matteo Rosoli Pitched at TRANSFORM 2026 in Berlin [Matteo Rosoli](https://www.linkedin.com/in/matteo-rosoli-9601a1266/), CEO and co-founder of eustella, delivered the first-ever public pitch of eustella at [TRANSFORM 2026](https://www.linkedin.com/company/transformshow/) in Berlin — one of Europe's most important events for media innovation, digital transformation, and emerging technology. The response was outstanding. eustella's vision for a sovereign, privacy-first AI agent built in Europe resonated strongly with an audience of media leaders, technology innovators, and investors who understand firsthand why European AI independence matters. ## The first public pitch TRANSFORM 2026 marked the moment eustella stepped onto the public stage. Matteo Rosoli presented the case for a European AI assistant that keeps data in Europe, runs on European and open-source models, and never trains on user conversations — a direct alternative to the US and Chinese AI products that dominate the market today. The pitch covered: - **The problem**: Every major AI chatbot — ChatGPT, Claude, Gemini, Grok, DeepSeek — stores European user data on foreign servers, trains on conversations by default, and is controlled by Big Tech or state-linked entities. - **The opportunity**: 133 million monthly active LLM users in Europe, but almost none of them are using a European product. The market gap is enormous. - **The solution**: eustella — a hyperpersonalised AI agent that is fully GDPR-compliant, aligned with the EU AI Act, and built by an independent European team with no Big Tech investors. ## Why TRANSFORM was the right stage TRANSFORM brings together the people shaping the future of media, technology, and digital business in Europe. The media and publishing industry understands better than most why data sovereignty matters. Newsrooms were among the first to feel the impact of Big Tech's data practices — from scraped content to opaque algorithmic distribution. eustella is built by [newsrooms.ai](https://newsrooms.ai), an independent European AI company founded in Vienna, Austria, by a team of AI experts and media professionals. The connection between media innovation and trustworthy AI is not incidental — it is the foundation of everything eustella stands for. Presenting at TRANSFORM was a natural fit, and the enthusiastic reception confirmed that the European market is ready for an AI product that puts privacy, sovereignty, and user trust first. ## What comes next The momentum from TRANSFORM 2026 fuels what comes next. The eustella beta launches on **15 April 2026**. Want to get involved? Reach out to Matteo Rosoli on [LinkedIn](https://www.linkedin.com/in/matteo-rosoli-9601a1266/) or sign up for early access at [eustella.com](https://eustella.com). [Join the waitlist →](/) --- # eustella — Podcast: \"There Is a Huge Market Gap in Europe for a Sovereign AI Agent\ URL: https://eustella.com/blog/europe-market-gap-sovereign-ai-agent Topic: podcast Last updated: 2026-03-19T00:00:00Z ## Podcast: \"There Is a Huge Market Gap in Europe for a Sovereign AI Agent\ *Originally published on [Trending Topics](https://www.trendingtopics.eu/openclaw-fuer-europa/) on 18 March 2026 by Jakob Steinschaden, co-written by newsrooms.ai.* "Europe consumes AI intensively, but primarily strengthens foreign ecosystems." That is the core finding of a new study on AI, startups, and Europe. Yet Europe has enormous potential for a real ChatGPT competitor — it lacks neither talent, nor a market, nor the building blocks. So why shouldn't a startup pick up where Peter Steinberger left off before he moved from Vienna to OpenAI? Jakob Steinschaden, co-founder of Trending Topics and newsrooms, and Matteo Rosoli, CEO of newsrooms, discuss: - **OpenClaw as a toolbox for AI models** — What Perplexity and others are building with it - **Peter Steinberger's move to OpenAI** — What the Viennese developer could achieve there - **Meta's $2 billion Manus acquisition** — Virtual machines and AI agents - **Chinese open-source models overtaking the West** — The rise of Kimi 2.5 and others - **Token costs compared** — Why €1,300 can be burned overnight - **Europe's AI gap** — Mistral, defense tech, and the lack of data sovereignty - **Vision for a European AI agent** — Proactive, connected, and privacy-first ## Why this matters for eustella eustella is being built to fill exactly the market gap described in this podcast. Europe has 450 million people, world-class AI regulation through GDPR and the AI Act, and a clear demand for an AI assistant that keeps data sovereign and operates under European law. eustella is that assistant — a personal AI agent built by newsrooms.ai in Vienna, designed to be proactive, autonomous, and fully compliant with European data protection standards. [Listen to the full podcast on Trending Topics →](https://www.trendingtopics.eu/openclaw-fuer-europa/) --- # eustella — eustella Launches Today: The Waitlist for Europe's AI Agent Is Open URL: https://eustella.com/blog/eustella-waitlist-launch Topic: launch Last updated: 2026-03-18T00:00:00Z ## eustella Launches Today: The Waitlist for Europe's AI Agent Is Open Today, 18 March 2026, eustella opens its waitlist. eustella is a personal AI agent built in Europe, for Europe — designed as the secure alternative to ChatGPT, Claude, Gemini, and every other AI assistant that trains on your data, stores it on US servers, or feeds it into an advertising machine. The waitlist is open now at [eustella.com](https://eustella.com). ## What is eustella? eustella is not another bland AI chatbot. eustella is a hyperpersonalised AI assistant — an agent that actually knows you, remembers your context, plans tasks on your behalf, and learns from every interaction to become more useful over time. eustella runs exclusively on European and open-source AI models. All data is stored and processed in EU data centres. No Big Tech investors, no military contracts, no hidden agendas. ## Why now? Every major AI chatbot — ChatGPT, Claude, Gemini, Grok, Copilot, Meta AI, Perplexity, DeepSeek — trains on your conversations by default. The opt-outs are buried, gated behind paywalls, or not retroactive. Your data sits on US or Chinese servers, subject to foreign laws that allow government access without European judicial oversight. Even Mistral, which calls itself European, is funded by Microsoft, NVIDIA, and US venture capital — and actively lobbied to weaken the EU AI Act. Europe deserves better. Europe deserves its own AI. ## What to expect from the beta The eustella beta launches on **15 April 2026**. Here is what we are building: ## Hyperpersonalised AI assistant No generic chatbot responses. eustella is an agent that knows you, remembers your preferences and history, can plan tasks autonomously, and learns from your interactions. The more you use eustella, the more useful it becomes — without ever selling or sharing what it learns. ## Everything you love about AI eustella can help you create, research, analyse, study, sort, refine — everything you are already used to from general AI chat apps. Writing, brainstorming, document analysis, code generation, summarisation, translation across 8 European languages. All the power, none of the data trade-offs. ## The first tool connections With the beta, eustella will ship its first tool integrations — connecting your AI assistant to the services and workflows you use every day. More details coming soon. ## After the beta: the full vision The beta is just the beginning. Our goal is a full personal AI assistant on your phone that proactively manages and helps you throughout your work and personal life — scheduling, reminders, research, communication, task management — all fully secure, all fully private, all fully European. An AI that works for you, not for advertisers. Not for governments. Not for Big Tech. ## Who builds eustella? eustella is built by [newsrooms.ai](https://newsrooms.ai), an independent European AI company based in Vienna, Austria. The founding team of AI experts and media professionals is committed to European digital sovereignty, privacy, and responsible AI. eustella is independently funded. No Big Tech investors. No conflicting interests. ## Join the waitlist Sign up today at [eustella.com](https://eustella.com) to secure your spot. Early access members will receive: - **Priority access** to the beta on 15 April 2026 - **Early-bird pricing** when eustella launches publicly - **Direct input** on features and roadmap The waitlist is free. Your data stays in Europe. [Sign up for early access →](/) --- # eustella — The Privacy Problem with AI Chatbots in 2026 URL: https://eustella.com/blog/ai-chatbots-privacy-problem-2026 Topic: privacy Last updated: 2026-03-17T00:00:00Z ## The Privacy Problem with AI Chatbots in 2026 Every major AI chatbot trains on your conversations by default. Most store your data on US or Chinese servers, subject to foreign government access. Several have signed military contracts. And the opt-outs — when they exist — are buried in settings, gated behind paid plans, or simply not retroactive. eustella analysed the data practices of the 11 most widely used AI chatbots. Here is what we found. ## ChatGPT: $200M Pentagon contract, €15M GDPR fine ChatGPT by OpenAI is the world's most popular AI assistant. By default, OpenAI uses your conversations to train its models. The opt-out is buried in settings and is not retroactive — data already submitted has been used. All data is stored on US servers routed through Microsoft Azure. In June 2025, OpenAI launched its government division with a [$200 million contract from the Pentagon's Chief Digital and AI Office (CDAO)](https://breakingdefense.com/2025/06/openai-for-government-launches-with-200m-win-from-pentagon-cdao/) [1]. Italy's data protection authority [fined OpenAI €15 million](https://www.euronews.com/next/2024/12/20/italys-privacy-watchdog-fines-openai-15-million-after-probe-into-chatgpt-data-collection) for GDPR violations in 2024 [2]. OpenAI is backed by Microsoft ($13 billion+). All data flows through Microsoft Azure. OpenAI is restructuring into a for-profit corporation. Note: OpenAI now offers [EU data residency for Enterprise and Education customers](https://openai.com/index/introducing-data-residency-in-europe/) since February 2025 [3], but free-tier user data is still processed in the United States. ## Claude: Trains on chats since September 2025 Claude by Anthropic is a capable AI assistant — but it is a US product, backed by Amazon ($8 billion) and Google ($3 billion+). Since [September 2025, Anthropic trains on your conversations by default](https://techcrunch.com/2025/08/28/anthropic-users-face-a-new-choice-opt-out-or-share-your-data-for-ai-training/) unless you opt out [4]. The opt-out interface has been [criticised as a dark pattern](https://the-decoder.com/anthropic-uses-a-questionable-dark-pattern-to-obtain-user-consent-for-ai-data-use-in-claude/) that defies GDPR guidelines [5]. Opted-in data may be retained for up to 5 years. Anthropic has no European offices and no EU data centres. All infrastructure runs on AWS and Google Cloud, and the US CLOUD Act applies — meaning the US government can compel data access. ## Google Gemini: 22 data categories, Pentagon AI for 3 million personnel Google Gemini collects 22 out of 35 possible data categories — more than any other AI chatbot. Your conversations are used for training by default, reviewed by human employees, and retained for up to 3 years even if you delete them. Gemini now powers the Pentagon's [GenAI.mil platform, rolling out to all 3 million military, civilian, and contractor personnel](https://breakingdefense.com/2025/12/pentagon-rolls-out-genai-platform-to-all-personnel-using-googles-gemini/) [6]. Gemini is part of Google — the world's largest advertising company. Your AI usage feeds the same ecosystem that tracks you across Search, Gmail, Maps, YouTube, Chrome, and Android. ## Grok: Owned by SpaceX, no opt-out since January 2026 Grok by xAI is the world's third-largest AI chatbot by web traffic. xAI is now a [wholly owned subsidiary of SpaceX](https://www.cnn.com/2026/02/02/tech/spacex-acquires-xai-elon-musk), controlled by Elon Musk, following its acquisition in February 2026 [7]. X's January 2026 terms of service grant a perpetual, worldwide license to use your content for AI training with no opt-out. Grok is under active GDPR investigation across multiple EU countries after [NOYB filed complaints in nine EU countries in August 2024](https://noyb.eu/en/twitters-ai-plans-hit-9-more-gdpr-complaints), alleging training on EU users' data without consent [8]. Musk led the DOGE initiative, which [sought access to US federal databases](https://www.npr.org/2025/02/04/nx-s1-5285403/musks-doge-group-has-access-to-the-federal-payments-system-what-does-that-mean) containing sensitive data on millions of citizens — Treasury payment systems were confirmed accessed, while IRS access was blocked by courts [9]. ## DeepSeek: All data in China, 1 million+ chat logs exposed DeepSeek is a powerful AI chatbot from China — but all your data is stored on servers in mainland China, subject to Chinese national security laws that compel companies to hand over data on government request. DeepSeek was [banned in Italy on 30 January 2025](https://www.euronews.com/next/2025/01/31/deepseek-ai-blocked-by-italian-authorities-as-others-member-states-open-probes) [10], is under investigation across the EU, and security researchers discovered an unprotected database with over 1 million lines of plaintext chat histories. The opt-out was only added after South Korean regulators intervened. ## Microsoft Copilot: $9B shared Pentagon contract, Recall screenshots your screen Microsoft Copilot is bundled across Windows, Office, Edge, and Bing. Consumer Copilot conversations are used to train AI models by default. Microsoft [paused this training in the EEA in August 2024](https://www.microsoft.com/en-us/microsoft-copilot/blog/2024/08/16/transparency-and-control-in-consumer-data-use/) rather than face GDPR enforcement — acknowledging the GDPR conflict rather than fixing the underlying practice [11]. Microsoft deploys GPT-4 in top-secret Pentagon clouds as part of the [Joint Warfighting Cloud Capability (JWCC) — a $9 billion contract shared across Microsoft, Amazon, Google, and Oracle](https://www.cnbc.com/2022/12/07/google-oracle-amazon-and-microsoft-awarded-9-billion-pentagon-cloud-deals.html) [12]. Consumer Copilot and military Azure operate on physically and logically separate infrastructure. The Windows Recall feature screenshots your screen every few seconds. The US House of Representatives banned its own staff from using Copilot over data security risks. ## Meta AI: €2.5B in GDPR fines, breaks WhatsApp encryption Meta AI is embedded across Facebook, Instagram, WhatsApp, and Messenger. Meta uses your AI conversations to personalise ads and train models. Invoking Meta AI in a WhatsApp chat breaks end-to-end encryption. Since May 2025, public posts from EU users are used for AI training by default. Meta has been fined over €2.5 billion under GDPR and paid [$5 billion to the US Federal Trade Commission](https://www.ftc.gov/news-events/news/press-releases/2019/07/ftc-imposes-5-billion-penalty-sweeping-new-privacy-restrictions-facebook) for the Cambridge Analytica scandal [13]. Meta is now an active defence contractor, [partnered with Anduril since May 2025](https://www.cnbc.com/2025/05/29/meta-anduril-partner-on-vr-ar-project-intended-for-us-army.html) for Pentagon AI and AR/VR projects [14]. ## Perplexity: Funded by Bezos, scrapes despite no-crawl rules Perplexity is a US-based AI search engine funded by Jeff Bezos, NVIDIA, and SoftBank, with a $750 million Microsoft Azure dependency. It trains on user data by default and has been caught scraping websites that explicitly blocked AI crawlers. Perplexity faces active lawsuits from The New York Times, Reddit, News Corp, and others. Cloudflare has publicly confirmed that Perplexity's bots ignore no-crawl rules. ## Mistral: Funded by Microsoft, lobbied against the EU AI Act Mistral positions itself as Europe's AI champion — but it is heavily funded by Microsoft, NVIDIA, and major US venture capital firms. Mistral trains on free-tier user data by default and was hit with a [GDPR complaint filed by French lawyer Jeremy Roche with CNIL in February 2025](https://sifted.eu/articles/mistral-privacy-policy-gdpr-news) for gating the opt-out behind paid plans [15]. Mistral actively lobbied to weaken the EU AI Act — [Corporate Europe Observatory documented](https://corporateeurope.org/en/2024/03/trojan-horses-how-european-startups-teamed-big-tech-gut-ai-act) how this effort aligned with Big Tech interests [16]. Mistral has a [direct military contract with the French armed forces](https://www.techrepublic.com/article/news-mistral-french-military-ai-deal/) [17], and indirect defence partnerships through Helsing (Germany) and Faculty AI (United Kingdom). ## Character.AI: Trains on intimate conversations, teen safety lawsuits Character.AI is the fifth-largest AI chatbot by web traffic. It trains on your conversations by default and uses chat data for targeted advertising since August 2025. Character.AI faces multiple lawsuits linked to teen suicides. Google invested $2.7 billion and is named as a co-defendant. In January 2026, Character.AI and Google agreed to mediate settlements — the first major legal settlement over AI-related harm. ## OpenClaw: 40,000+ exposed instances, malicious plugins OpenClaw is a popular open-source AI agent with over 316,000 GitHub stars. It can control your computer, manage your calendar, send emails, and run shell commands. But it is insecure by default — security researchers initially found over 40,000 exposed instances on the public internet, many without authentication; later scans revealed between 135,000 and 220,000+ vulnerable deployments. 12% of community-built skills on ClawHub were found to be malicious, a figure that has since grown to approximately 20%. OpenClaw stores API keys in plaintext and has no built-in encryption. Its creator joined OpenAI in February 2026, leaving the project's governance uncertain. ## The pattern is clear Every major AI chatbot — American, Chinese, and even nominally European — trains on your data by default, stores it outside the EU, or is funded by the same Big Tech companies that profit from your information. eustella is designed as the European alternative. Your data stays in Europe, is never sold, and is never used for training. eustella runs exclusively on European and open-source AI models, with no Big Tech investors, no military contracts, and no hidden agendas. [Sign up for early access to eustella →](/) --- ## Sources 1. Breaking Defense. "OpenAI for Government launches with $200M win from Pentagon CDAO." June 2025. [Link](https://breakingdefense.com/2025/06/openai-for-government-launches-with-200m-win-from-pentagon-cdao/) 2. Euronews. "Italy's privacy watchdog fines OpenAI 15 million after probe into ChatGPT data collection." December 2024. [Link](https://www.euronews.com/next/2024/12/20/italys-privacy-watchdog-fines-openai-15-million-after-probe-into-chatgpt-data-collection) 3. OpenAI. "Introducing data residency in Europe." February 2025. [Link](https://openai.com/index/introducing-data-residency-in-europe/) 4. TechCrunch. "Anthropic users face a new choice: opt out or share your data for AI training." August 2025. [Link](https://techcrunch.com/2025/08/28/anthropic-users-face-a-new-choice-opt-out-or-share-your-data-for-ai-training/) 5. The Decoder. "Anthropic uses a questionable dark pattern to obtain user consent for AI data use in Claude." 2025. [Link](https://the-decoder.com/anthropic-uses-a-questionable-dark-pattern-to-obtain-user-consent-for-ai-data-use-in-claude/) 6. Breaking Defense. "Pentagon rolls out GenAI platform to all personnel, using Google's Gemini." December 2025. [Link](https://breakingdefense.com/2025/12/pentagon-rolls-out-genai-platform-to-all-personnel-using-googles-gemini/) 7. CNN. "SpaceX acquires xAI." February 2026. [Link](https://www.cnn.com/2026/02/02/tech/spacex-acquires-xai-elon-musk) 8. NOYB. "Twitter's AI plans hit with 9 more GDPR complaints." August 2024. [Link](https://noyb.eu/en/twitters-ai-plans-hit-9-more-gdpr-complaints) 9. NPR. "Musk's DOGE group has access to the federal payments system. What does that mean?" February 2025. [Link](https://www.npr.org/2025/02/04/nx-s1-5285403/musks-doge-group-has-access-to-the-federal-payments-system-what-does-that-mean) 10. Euronews. "DeepSeek AI blocked by Italian authorities." January 2025. [Link](https://www.euronews.com/next/2025/01/31/deepseek-ai-blocked-by-italian-authorities-as-others-member-states-open-probes) 11. Microsoft. "Transparency and control in consumer data use." August 2024. [Link](https://www.microsoft.com/en-us/microsoft-copilot/blog/2024/08/16/transparency-and-control-in-consumer-data-use/) 12. CNBC. "Google, Oracle, Amazon and Microsoft awarded $9 billion Pentagon cloud deals." December 2022. [Link](https://www.cnbc.com/2022/12/07/google-oracle-amazon-and-microsoft-awarded-9-billion-pentagon-cloud-deals.html) 13. FTC. "FTC imposes $5 billion penalty, sweeping new privacy restrictions on Facebook." July 2019. [Link](https://www.ftc.gov/news-events/news/press-releases/2019/07/ftc-imposes-5-billion-penalty-sweeping-new-privacy-restrictions-facebook) 14. CNBC. "Meta, Anduril partner on VR/AR project intended for US Army." May 2025. [Link](https://www.cnbc.com/2025/05/29/meta-anduril-partner-on-vr-ar-project-intended-for-us-army.html) 15. Sifted. "Mistral privacy policy GDPR complaint." February 2025. [Link](https://sifted.eu/articles/mistral-privacy-policy-gdpr-news) 16. Corporate Europe Observatory. "Trojan horses: how European startups teamed up with Big Tech to gut the AI Act." March 2024. [Link](https://corporateeurope.org/en/2024/03/trojan-horses-how-european-startups-teamed-big-tech-gut-ai-act) 17. TechRepublic. "Mistral French military AI deal." 2025. [Link](https://www.techrepublic.com/article/news-mistral-french-military-ai-deal/) --- # eustella — Why Europe Needs Its Own AI URL: https://eustella.com/blog/why-europe-needs-its-own-ai Topic: european-ai Last updated: 2026-03-15T00:00:00Z ## Why Europe Needs Its Own AI Europe is the world's most advanced regulatory environment for artificial intelligence. The EU AI Act, GDPR, and the Digital Services Act create a framework that no other region matches. But regulation alone is not enough. Europe also needs its own AI products — built here, hosted here, governed by European law. ## The numbers speak for themselves According to the [State of AI in Europe](https://www.prosus.com/state-of-ai-europe) report by Prosus and Dealroom, Europe's AI ecosystem is brimming with potential — but the value it creates overwhelmingly flows elsewhere: - **325,000+ AI professionals** work across Europe — though 53% are employed in traditional sectors like consulting, industrial, and finance, with only 33% working in tech companies [1] - **133 million monthly active LLM users in Europe** — but they are overwhelmingly using American products like ChatGPT, not European ones. As the report puts it: "Europeans consume AI brilliantly, but train the algorithms owned by others" [1] - **European founders are behind 25% of all global AI unicorns** — but many of these founders relocated to the US to build their companies, draining European talent rather than reinforcing the local ecosystem [1] - **~900 new AI startups** are founded in Europe every year [1] The picture is clear: Europe does not lack talent or demand. What it lacks is a consumer-facing AI product that competes with ChatGPT, Claude, or Gemini — while keeping data and value in Europe. ## The problem with US and Chinese AI Every major AI chatbot is either American or Chinese. This creates three fundamental problems for European users: ## 1. Data leaves Europe When you use ChatGPT, your conversations are stored on US servers. When you use DeepSeek, they go to China. In both cases, your data is subject to foreign laws that can override European privacy protections. The US CLOUD Act [2] compels American companies to hand over data stored anywhere in the world when served with a lawful warrant or subpoena — regardless of where the data is physically located and without requiring European judicial oversight. Chinese national security legislation goes further. The National Intelligence Law (2017) requires all organisations and citizens to "support, assist and cooperate with national intelligence work" (Article 7) [3]. The Cybersecurity Law (2017, Article 28) and Data Security Law (2021) impose similar obligations [4]. ## 2. Your data trains their models ChatGPT, Claude, Gemini, Grok, Copilot, Meta AI, Perplexity, Mistral, Character.AI, and DeepSeek all use your conversations to train their models by default. The opt-outs — when they exist — are buried in settings, gated behind paid plans, or not retroactive. ## 3. Big Tech controls the stack OpenAI is backed by Microsoft. Claude runs on Amazon and Google infrastructure. Grok is owned by SpaceX. Meta AI feeds the advertising machine. Even Mistral, which positions itself as European, is funded by Microsoft, NVIDIA, and US venture capital. The companies building AI are the same companies that profit from harvesting your data. ## What Europe has that others don't Europe's regulatory framework is not a weakness — it is a competitive advantage: - **GDPR** gives Europeans the right to know, access, correct, and delete their data. It requires explicit consent for data processing and imposes real penalties for violations. - **The EU AI Act** is the world's first comprehensive AI regulation, with a phased rollout from 2024 to full applicability in August 2027 [5]. It classifies AI systems by risk level and bans unacceptable practices like social scoring and real-time biometric surveillance in public spaces — with narrow exceptions only for terrorism threats, missing persons, and serious crime [6]. - **Data sovereignty** means your data is stored in the EU, processed under EU law, and not subject to foreign government access. These are not obstacles to innovation. They are the foundation of trustworthy AI. ## eustella: AI built for Europeans eustella is a full-featured AI assistant built in Europe, for Europe. It handles the same core tasks as ChatGPT — answering questions, writing texts, analysing documents, generating code, brainstorming ideas — with one fundamental difference: your data is never sold, never shared, and never used to train AI models. eustella runs exclusively on European and open-source AI models. No Big Tech data harvesters from the US or China. No military partnerships. No hidden agendas. Just a powerful AI assistant that puts you first. ## Key facts about eustella - **European hosting**: All data stored and processed in EU data centres - **No training on user data**: Your conversations are yours - **Open-source models**: Transparent, auditable, independent - **GDPR compliant by design**: Privacy is the foundation, not a feature - **Mobile-first**: Designed for smartphones, not just desktops - **8 European languages**: English, German, French, Spanish, Italian, Dutch, Portuguese, Polish ## The time is now The European AI ecosystem has the talent, the funding, and the regulatory advantage. What it needs is products that Europeans can trust — products that respect their data, their rights, and their values. eustella is that product. [Sign up for early access →](/) --- ## Sources 1. Prosus & Dealroom, "State of AI in Europe" report — [prosus.com/state-of-ai-europe](https://www.prosus.com/state-of-ai-europe) 2. US CLOUD Act (Clarifying Lawful Overseas Use of Data Act), 2018 — requires US companies to comply with lawful warrants/subpoenas for data regardless of storage location 3. National Intelligence Law of the People's Republic of China, 2017, Article 7 — [chinalawtranslate.com](https://www.chinalawtranslate.com/en/national-intelligence-law-of-the-p-r-c-2017/) 4. Cybersecurity Law of the People's Republic of China, 2017, Article 28; Data Security Law, 2021 5. EU AI Act implementation timeline — [artificialintelligenceact.eu/implementation-timeline](https://artificialintelligenceact.eu/implementation-timeline/) 6. EU AI Act, Article 5 (prohibited practices and exceptions) — [artificialintelligenceact.eu/article/5](https://artificialintelligenceact.eu/article/5/) --- # eustella — GDPR and AI Chatbots: What You Need to Know in 2026 URL: https://eustella.com/blog/gdpr-ai-chatbots-what-you-need-to-know Topic: gdpr Last updated: 2026-03-12T00:00:00Z ## GDPR and AI Chatbots: What You Need to Know in 2026 The General Data Protection Regulation (GDPR) is the world's strongest data protection law. But AI chatbots are testing its limits. Every major AI assistant collects conversational data, and most use it for model training — often without meaningful consent. Here is what GDPR says, what companies are doing, and how to protect yourself. ## What GDPR requires GDPR establishes clear rules for how personal data must be handled: - **Lawful basis**: Companies need a valid legal reason to process your data — usually consent or legitimate interest - **Purpose limitation**: Data collected for one purpose cannot be used for another without consent - **Data minimisation**: Companies should collect only the data they need - **Right to access**: You can request a copy of all data a company holds about you - **Right to erasure**: You can request deletion of your data (the "right to be forgotten") - **Right to object**: You can object to automated processing, including profiling - **Data portability**: You can request your data in a machine-readable format - **Breach notification**: Companies must report data breaches to the supervisory authority within 72 hours ([Article 33](https://gdpr-info.eu/art-33-gdpr/)). Data subjects must be informed without undue delay when the breach poses a high risk ([Article 34](https://gdpr-info.eu/art-34-gdpr/)) Violations carry two tiers of fines. The upper tier ([Art. 83(5)](https://gdpr-info.eu/art-83-gdpr/)) covers core violations — up to 4% of global annual revenue or €20 million, whichever is higher. The lower tier ([Art. 83(4)](https://gdpr-info.eu/art-83-gdpr/)) covers administrative and technical obligations — up to 2% of global annual revenue or €10 million. ## How AI chatbots violate GDPR Despite these protections, AI chatbot companies have repeatedly violated GDPR: ## Training on user data without valid consent ChatGPT, Claude, Gemini, Grok, Copilot, Meta AI, Perplexity, Mistral, and Character.AI all train on user conversations by default. Under GDPR, using personal data for a new purpose (AI training) requires separate, informed, freely given consent. Burying an opt-out in settings does not meet this standard. ## Dark patterns in opt-out interfaces Anthropic (Claude) was [criticised for introducing an opt-out interface](https://the-decoder.com/anthropic-uses-a-questionable-dark-pattern-to-obtain-user-consent-for-ai-data-use-in-claude/) that privacy advocates described as a "dark pattern" — designed to discourage users from opting out. Under GDPR, withdrawing consent must be as easy as giving it. ## Gating privacy behind paywalls Mistral initially made the training opt-out available only on paid plans. A [GDPR complaint was filed by French lawyer Jeremy Roche with CNIL](https://sifted.eu/articles/mistral-privacy-policy-gdpr-news) in February 2025 because Article 12 requires that privacy controls be freely accessible regardless of payment tier. ## Notable GDPR enforcement actions against AI companies | Company | Fine/Action | Year | |---------|------------|------| | Meta | [~€2.5 billion in cumulative GDPR fines](https://www.enforcementtracker.com/) (including €1.2B for data transfers, €390M for behavioural advertising, €251M in 2024, €225M for WhatsApp) | 2021-2025 | | OpenAI (ChatGPT) | [€15 million fine (Italy)](https://www.euronews.com/next/2024/12/20/italys-privacy-watchdog-fines-openai-15-million-after-probe-into-chatgpt-data-collection) | 2024 | | DeepSeek | [Banned in Italy](https://www.euronews.com/next/2025/01/31/deepseek-ai-blocked-by-italian-authorities-as-others-member-states-open-probes) (January 30, 2025), under investigation across EU | 2025 | | Grok/X | [NOYB complaints filed across 9 EU countries](https://noyb.eu/en/twitters-ai-plans-hit-9-more-gdpr-complaints) | August 2024 | | Mistral | [GDPR complaint for paywall-gated opt-out](https://sifted.eu/articles/mistral-privacy-policy-gdpr-news) (filed with CNIL) | February 2025 | | Microsoft | [Paused Copilot training in EEA](https://www.microsoft.com/en-us/microsoft-copilot/blog/2024/08/16/transparency-and-control-in-consumer-data-use/) to avoid enforcement | August 2024 | ## The EU AI Act adds a new layer The [EU AI Act](https://artificialintelligenceact.eu/implementation-timeline/), which entered into force in August 2024 and is being implemented in stages through 2027, complements GDPR with AI-specific rules. Key dates: prohibited practices took effect February 2025, general-purpose AI (GPAI) rules apply from August 2025, high-risk system requirements from August 2026, and full applicability from August 2027. - **Risk classification**: AI systems are classified as minimal, limited, high, or unacceptable risk - **Transparency requirements**: AI systems must disclose when content is AI-generated - **Prohibited practices**: Social scoring, manipulative AI techniques, and [real-time biometric surveillance in public spaces](https://artificialintelligenceact.eu/article/5/) are banned, with narrow exceptions for terrorism threats, missing persons searches, and serious crime investigations - **Foundation model obligations**: Providers of general-purpose AI models must document training data (a [mandatory training data summary template](https://www.wilmerhale.com/en/insights/blogs/wilmerhale-privacy-and-cybersecurity-law/european-commission-releases-mandatory-template-for-public-disclosure-of-ai-training-data) was published in July 2025), comply with copyright law, and publish [energy consumption data](https://www.whitecase.com/insight-alert/energy-efficiency-requirements-under-eu-ai-act) Together, GDPR and the AI Act create the most comprehensive AI governance framework in the world. ## What to look for in a privacy-respecting AI assistant When choosing an AI chatbot, ask these questions: 1. **Where is my data stored?** Look for EU-only data centres. Avoid services that route data through US or Chinese servers. 2. **Is my data used for training?** Read the privacy policy. If the default is opt-in to training, the company profits from your data. 3. **Who funds the company?** Big Tech investors (Microsoft, Amazon, Google) create conflicts of interest. An AI company funded by the same companies that profit from your data cannot be truly independent. 4. **Is there a meaningful opt-out?** An opt-out buried in settings, gated behind payment, or not retroactive is not meaningful consent. 5. **Does the company have EU presence?** Companies without European offices or data centres are harder to hold accountable under GDPR. ## eustella: GDPR compliant by design eustella is designed from the ground up to comply with GDPR and the EU AI Act: - **All data stored in EU data centres** — no transfers to the US, China, or any third country - **No training on user data** — your conversations are never used to improve models - **No data selling or sharing** — your data is yours - **European and open-source models only** — transparent, auditable, independent from Big Tech - **Built by an independent European company** — AI Newsrooms Technology GmbH, Vienna, Austria eustella does not need an opt-out because it never opts you in. [Sign up for early access to eustella →](/) --- ## Sources - **GDPR Articles 33 & 34** (breach notification): [gdpr-info.eu](https://gdpr-info.eu/art-33-gdpr/) - **GDPR Article 83** (fine tiers): [gdpr-info.eu](https://gdpr-info.eu/art-83-gdpr/) - **OpenAI €15M fine (Italy, 2024)**: [Euronews](https://www.euronews.com/next/2024/12/20/italys-privacy-watchdog-fines-openai-15-million-after-probe-into-chatgpt-data-collection) - **DeepSeek banned in Italy (January 2025)**: [Euronews](https://www.euronews.com/next/2025/01/31/deepseek-ai-blocked-by-italian-authorities-as-others-member-states-open-probes) - **Grok/X NOYB complaints (August 2024)**: [noyb.eu](https://noyb.eu/en/twitters-ai-plans-hit-9-more-gdpr-complaints) - **Mistral GDPR complaint (February 2025)**: [Sifted](https://sifted.eu/articles/mistral-privacy-policy-gdpr-news) - **Microsoft Copilot training pause (August 2024)**: [Microsoft Blog](https://www.microsoft.com/en-us/microsoft-copilot/blog/2024/08/16/transparency-and-control-in-consumer-data-use/) - **Anthropic dark pattern criticism**: [The Decoder](https://the-decoder.com/anthropic-uses-a-questionable-dark-pattern-to-obtain-user-consent-for-ai-data-use-in-claude/) - **EU AI Act implementation timeline**: [artificialintelligenceact.eu](https://artificialintelligenceact.eu/implementation-timeline/) - **EU AI Act Article 5** (prohibited practices): [artificialintelligenceact.eu](https://artificialintelligenceact.eu/article/5/) - **AI training data summary template (July 2025)**: [WilmerHale](https://www.wilmerhale.com/en/insights/blogs/wilmerhale-privacy-and-cybersecurity-law/european-commission-releases-mandatory-template-for-public-disclosure-of-ai-training-data) - **GPAI energy consumption requirements**: [White & Case](https://www.whitecase.com/insight-alert/energy-efficiency-requirements-under-eu-ai-act) --- # eustella — OpenClaw's Security Problem: Hundreds of Thousands of Exposed Instances and Malicious Plugins URL: https://eustella.com/blog/openclaw-security-risks Topic: openclaw Last updated: 2026-03-10T00:00:00Z ## OpenClaw's Security Problem: Hundreds of Thousands of Exposed Instances and Malicious Plugins OpenClaw is one of the most popular open-source projects in the world, with [over 300,000 GitHub stars](https://www.star-history.com/blog/openclaw-surpasses-react-most-starred-software) — surpassing React, Linux, and nearly every other repository on the platform. It can control your computer, manage your calendar, send emails, browse the web, and run shell commands. OpenClaw represents the future of AI — autonomous agents that act on your behalf. But OpenClaw has a serious security problem. ## Over 135,000 exposed instances SecurityScorecard initially discovered [over 40,000 OpenClaw instances](https://www.infosecurity-magazine.com/news/researchers-40000-exposed-openclaw/) exposed on the public internet. Subsequent research by Censys and Bitdefender found the problem was far worse: [over 135,000 exposed instances across 82 countries](https://www.bitdefender.com/en-us/blog/hotforsecurity/135k-openclaw-ai-agents-exposed-online), with some estimates exceeding 220,000. Researchers found that 63% of observed deployments were vulnerable. OpenClaw binds to `0.0.0.0:18789` (all network interfaces) by default rather than localhost, and lacks rate limits — meaning many instances are [effectively exposed](https://docs.openclaw.ai/gateway/security) even when the Gateway's token-based authentication is present. Many users never change the defaults. ## CVE-2026-25253: One-click remote code execution In early 2026, researchers disclosed [CVE-2026-25253](https://nvd.nist.gov/vuln/detail/CVE-2026-25253), a critical vulnerability (CVSS 8.8) that allowed one-click remote code execution via crafted URLs. An attacker could send a link that, when clicked by someone running OpenClaw, would execute arbitrary commands on their machine. The vulnerability was patched in v2026.1.29, but many instances remain unpatched. ## 12% to 20% of community plugins are malicious The ClawHub marketplace allows developers to share plugins (called "skills") that extend OpenClaw's capabilities. Koi Security [audited 2,857 ClawHub skills](https://dev.to/tiamatenity/openclaw-skill-malware-audit-341-malicious-skills-infecting-clawhub-cfk) and found 341 malicious entries — roughly 12% — containing data exfiltration, cryptocurrency miners, and backdoor access. As ClawHub grew to over 10,700 skills by mid-February 2026, the problem scaled with it. Bitdefender [identified 824+ malicious skills](https://particula.tech/blog/openclaw-security-crisis-malicious-ai-agents), pushing the malicious rate to approximately 20%. OpenClaw has no code signing, no sandboxing, and no mandatory review process for community plugins. ## Plaintext credentials OpenClaw stores API keys, passwords, and authentication tokens in plaintext configuration files at `~/.openclaw/agents/*/agent/auth-profiles.json`. If an attacker gains access to the instance, they get [every credential the agent has been configured with](https://github.com/openclaw/openclaw/issues/7139). ## No encryption at rest or in transit By default, OpenClaw does not encrypt data at rest. Communication between the agent and LLM providers is encrypted via HTTPS, but local storage, logs, and conversation history are unprotected. ## The Moltbook incident The ecosystem around OpenClaw has its own security problems. Moltbook, a social network for OpenClaw agents, [leaked 35,000 email addresses and 1.5 million API tokens](https://www.wiz.io/blog/exposed-moltbook-database-reveals-millions-of-api-keys) through an exposed Supabase database — giving attackers direct access to users' LLM provider accounts and billing. ## Why this matters OpenClaw is not a chatbot. It is an AI agent — software that takes actions on your behalf. When an agent is compromised, the attacker doesn't just read your messages. They can: - **Send emails as you** — phishing, impersonation, fraud - **Access your files** — documents, photos, credentials - **Execute system commands** — install malware, exfiltrate data, delete files - **Control connected services** — calendar, task managers, smart home devices The attack surface of an AI agent is fundamentally larger than a chatbot. Security cannot be an afterthought. ## OpenClaw's governance problem OpenClaw's original creator, Peter Steinberger, [joined OpenAI on February 15, 2026](https://techcrunch.com/2026/02/15/openclaw-creator-peter-steinberger-joins-openai/). Sam Altman announced the hire personally. Steinberger, previously the founder of PSPDFKit (sold for approximately $116M in 2021), was [one of Austria's most prominent tech founders](https://www.euronews.com/next/2026/02/16/austrian-creator-of-viral-openclaw-joins-openai-to-build-next-generation-of-ai-agents). The project was [moved to an independent open-source foundation](https://finance.yahoo.com/news/openclaw-founder-steinberger-joins-openai-223554158.html) backed by OpenAI, but its future direction and governance remain uncertain. When a project's lead moves to one of the largest AI companies in the world — and that same company funds the foundation — questions about independence, roadmap, and long-term maintenance are legitimate. ## eustella: The secure alternative eustella gives you the same agentic capabilities as OpenClaw — task execution, workflow automation, communication, app integrations — but secure, private, and hosted entirely in Europe. ## How eustella addresses OpenClaw's security gaps | Issue | OpenClaw | eustella | |-------|----------|---------| | Authentication | Binds to all interfaces by default | Required, multi-factor | | Data encryption | Plaintext storage | Encrypted at rest and in transit | | Plugin security | No review, 12–20% malicious | Curated, sandboxed, reviewed | | Credential storage | Plaintext config files | Encrypted secrets management | | Hosting | Self-hosted (user responsible) | Managed EU data centres | | Data sovereignty | Depends on LLM provider | All data stays in the EU | | Vulnerability response | User must patch manually | Managed, automatic updates | eustella is designed for people who want the power of an AI agent without the responsibility of securing one themselves. No exposed databases, no malicious plugins, no plaintext credentials. [Sign up for early access to eustella →](/) --- ## Sources 1. Star History Blog — ["OpenClaw surpasses React as most-starred software project"](https://www.star-history.com/blog/openclaw-surpasses-react-most-starred-software) 2. Infosecurity Magazine — ["Researchers find 40,000 exposed OpenClaw instances"](https://www.infosecurity-magazine.com/news/researchers-40000-exposed-openclaw/) 3. Bitdefender — ["135K OpenClaw AI agents exposed online"](https://www.bitdefender.com/en-us/blog/hotforsecurity/135k-openclaw-ai-agents-exposed-online) 4. OpenClaw Gateway Security Documentation — [docs.openclaw.ai/gateway/security](https://docs.openclaw.ai/gateway/security) 5. NVD — [CVE-2026-25253](https://nvd.nist.gov/vuln/detail/CVE-2026-25253) (CVSS 8.8, 1-click RCE) 6. Tiamatenity / Dev.to — ["OpenClaw skill malware audit: 341 malicious skills infecting ClawHub"](https://dev.to/tiamatenity/openclaw-skill-malware-audit-341-malicious-skills-infecting-clawhub-cfk) 7. Particula Tech — ["OpenClaw security crisis: Malicious AI agents"](https://particula.tech/blog/openclaw-security-crisis-malicious-ai-agents) 8. GitHub Issue #7139 — [Plaintext credential storage in auth-profiles.json](https://github.com/openclaw/openclaw/issues/7139) 9. Wiz Research — ["Exposed Moltbook database reveals millions of API keys"](https://www.wiz.io/blog/exposed-moltbook-database-reveals-millions-of-api-keys) 10. TechCrunch — ["OpenClaw creator Peter Steinberger joins OpenAI"](https://techcrunch.com/2026/02/15/openclaw-creator-peter-steinberger-joins-openai/) 11. Euronews — ["Austrian creator of viral OpenClaw joins OpenAI"](https://www.euronews.com/next/2026/02/16/austrian-creator-of-viral-openclaw-joins-openai-to-build-next-generation-of-ai-agents) 12. Yahoo Finance — ["OpenClaw founder Steinberger joins OpenAI, project moves to foundation"](https://finance.yahoo.com/news/openclaw-founder-steinberger-joins-openai-223554158.html) --- # eustella — Which AI Companies Have Military Contracts? URL: https://eustella.com/blog/military-contracts-ai-companies Topic: military Last updated: 2026-03-08T00:00:00Z ## Which AI Companies Have Military Contracts? The line between consumer AI and military AI is disappearing. The same companies that build the chatbots you use every day are signing billion-dollar contracts with defence ministries. Here is a factual overview of which AI companies have military ties — and what that means for your data. ## OpenAI (ChatGPT) OpenAI signed a [**$200 million contract**](https://breakingdefense.com/2025/06/openai-for-government-launches-with-200m-win-from-pentagon-cdao/) with the US Department of Defense in June 2025, awarded through the Pentagon's Chief Digital and Artificial Intelligence Office (CDAO). This is a significant shift — OpenAI's usage policy originally [prohibited military applications](https://theintercept.com/2024/01/12/open-ai-military-ban-chatgpt/). The company quietly removed this restriction on January 10, 2024. All ChatGPT data [flows through Microsoft Azure](https://dgtlinfra.com/chatgpt-openai-azure-cloud/), the same cloud provider used for classified Pentagon systems. ## Google (Gemini) Google Gemini now [**powers the Pentagon's GenAI.mil platform for all 3 million defence personnel**](https://breakingdefense.com/2025/12/pentagon-rolls-out-genai-platform-to-all-personnel-using-googles-gemini/). Google has a long history of military AI work, including the controversial Project Maven — a program for [analyzing drone imagery through object labeling](https://www.pbs.org/newshour/show/amid-pressure-from-employees-google-drops-pentagons-project-maven-account) — which led to approximately 4,000 employees signing a petition and around 12 resignations in 2018. Google did not renew the contract when it expired in March 2019. Despite those earlier promises to limit military AI, Google has steadily expanded its defence contracts. ## Microsoft (Copilot) Microsoft is one of four vendors (alongside Amazon, Google, and Oracle) sharing the Pentagon's [**$9 billion JWCC cloud contract**](https://www.cnbc.com/2022/12/07/google-oracle-amazon-and-microsoft-awarded-9-billion-pentagon-cloud-deals.html). Microsoft also won the IVAS augmented reality headset contract for the US Army, worth up to $21.88 billion — though the company [announced in February 2025](https://breakingdefense.com/2025/02/microsoft-announces-plan-to-slide-22-billion-ivas-contract-over-to-anduril/) that it is transferring the program to Anduril Industries. Microsoft [deploys GPT-4 in top-secret Pentagon cloud environments](https://defensescoop.com/2024/05/07/gpt-4-pentagon-azure-top-secret-cloud-microsoft/) via its classified Azure infrastructure. Consumer Copilot runs on Microsoft's commercial Azure cloud, which is [physically and logically separate](https://learn.microsoft.com/en-us/microsoft-copilot-studio/requirements-licensing-gcc) from its government and military infrastructure (Azure Government, Azure Government Top Secret). The two are distinct systems, but both are operated by the same company. ## Meta (Meta AI) Meta is entering the defence space through a [**partnership with Anduril Industries**](https://www.cnbc.com/2025/05/29/meta-anduril-partner-on-vr-ar-project-intended-for-us-army.html), announced in May 2025, to develop mixed-reality and XR devices (the EagleEye system) for the US Army's SBMC Next program. Separately, Meta [made its Llama models available](https://techcrunch.com/2024/11/04/meta-says-its-making-its-llama-models-available-for-us-national-security-applications/) for US national security applications in November 2024. This is a newer development — Meta historically avoided military work. The shift aligns with Meta's pivot toward AI-powered products and the lucrative US defence market. ## Mistral Despite positioning itself as Europe's AI champion, Mistral has expanding military ties across multiple countries. In January 2026, Mistral signed a [**direct framework agreement with the French Ministry of Defense**](https://www.techrepublic.com/article/news-mistral-french-military-ai-deal/). In Germany, Mistral has [partnered with Helsing](https://sifted.eu/articles/mistral-helsing-defence-ai-action-summit-paris), a defence technology company, rather than contracting directly with the government. In the UK, Mistral works with Faculty AI, a company that contracts with the UK Ministry of Defence. Mistral also lobbied to weaken the EU AI Act — the regulation designed to protect European citizens from AI risks. [Corporate Europe Observatory](https://corporateeurope.org/en/2024/03/trojan-horses-how-european-startups-teamed-big-tech-gut-ai-act) documented how this lobbying aligned with Big Tech interests. ## xAI/SpaceX xAI was [**acquired by SpaceX**](https://www.cnn.com/2026/02/02/tech/spacex-acquires-xai-elon-musk) in February 2026, creating a combined entity valued at approximately $1.25 trillion. SpaceX holds roughly [**$22 billion in total government contracts**](https://www.washingtonpost.com/technology/interactive/2025/elon-musk-business-government-contracts-funding/) ($13 billion+ from NASA, $5–8 billion+ from the Department of Defense). Elon Musk's DOGE initiative [sought access to US federal databases](https://www.npr.org/2025/02/04/nx-s1-5285403/musks-doge-group-has-access-to-the-federal-payments-system-what-does-that-mean), with confirmed access to Treasury payment systems and OPM personnel data, though IRS access was blocked or contested. The interconnection between X/Twitter, SpaceX, Tesla, and xAI creates a data ecosystem with direct ties to US government systems. ## What military contracts mean for your data Military contracts matter for consumer AI users because: 1. **Infrastructure overlap**: The same companies that process your conversations also serve classified military workloads. Even when infrastructure is separated, the corporate entity — and its legal obligations — are the same. 2. **Legal obligations**: Companies with military contracts are subject to government oversight, including the ability to compel data access under national security provisions. 3. **Mission drift**: A company that serves both consumers and militaries has conflicting priorities. Consumer privacy may be deprioritised when national security contracts are at stake. 4. **Normalisation**: When AI companies treat military applications as routine business, the ethical boundaries around AI use erode. ## Companies without military contracts Not all AI companies have military ties. Notable exceptions include: - **eustella** — Independent European AI company with no military contracts, no Big Tech investors, and a clear commitment to civilian applications only - Some smaller open-source AI labs and research organisations However, the trend in the industry is clearly toward increasing military engagement. ## eustella's position eustella is built by newsrooms.ai, an independent European AI company. eustella has no military contracts, no partnerships with defence ministries, and no plans to pursue them. eustella exists to serve individual users — not governments, not militaries, not advertisers. This is a deliberate choice. eustella believes that the same AI used for personal communication, creative work, and daily tasks should not be entangled with weapons systems and surveillance infrastructure. [Sign up for early access to eustella →](/) --- ## Sources - Breaking Defense — [OpenAI for Government launches with $200M win from Pentagon CDAO](https://breakingdefense.com/2025/06/openai-for-government-launches-with-200m-win-from-pentagon-cdao/) (June 2025) - The Intercept — [OpenAI quietly deletes ban on military use of ChatGPT](https://theintercept.com/2024/01/12/open-ai-military-ban-chatgpt/) (January 2024) - DGTL Infra — [ChatGPT and OpenAI on Azure Cloud](https://dgtlinfra.com/chatgpt-openai-azure-cloud/) - Breaking Defense — [Pentagon rolls out GenAI platform to all personnel using Google's Gemini](https://breakingdefense.com/2025/12/pentagon-rolls-out-genai-platform-to-all-personnel-using-googles-gemini/) (December 2025) - PBS NewsHour — [Google drops Pentagon's Project Maven account](https://www.pbs.org/newshour/show/amid-pressure-from-employees-google-drops-pentagons-project-maven-account) (2018) - CNBC — [Google, Oracle, Amazon and Microsoft awarded $9 billion Pentagon cloud deals](https://www.cnbc.com/2022/12/07/google-oracle-amazon-and-microsoft-awarded-9-billion-pentagon-cloud-deals.html) (December 2022) - Breaking Defense — [Microsoft announces plan to transfer $22 billion IVAS contract to Anduril](https://breakingdefense.com/2025/02/microsoft-announces-plan-to-slide-22-billion-ivas-contract-over-to-anduril/) (February 2025) - DefenseScoop — [GPT-4 deployed in Pentagon's Azure top-secret cloud](https://defensescoop.com/2024/05/07/gpt-4-pentagon-azure-top-secret-cloud-microsoft/) (May 2024) - Microsoft Learn — [Copilot Studio requirements and licensing for GCC](https://learn.microsoft.com/en-us/microsoft-copilot-studio/requirements-licensing-gcc) - CNBC — [Meta and Anduril partner on VR/AR project for US Army](https://www.cnbc.com/2025/05/29/meta-anduril-partner-on-vr-ar-project-intended-for-us-army.html) (May 2025) - TechCrunch — [Meta makes Llama available for US national security applications](https://techcrunch.com/2024/11/04/meta-says-its-making-its-llama-models-available-for-us-national-security-applications/) (November 2024) - TechRepublic — [Mistral signs French military AI deal](https://www.techrepublic.com/article/news-mistral-french-military-ai-deal/) (January 2026) - Sifted — [Mistral and Helsing defence AI partnership](https://sifted.eu/articles/mistral-helsing-defence-ai-action-summit-paris) - Corporate Europe Observatory — [How European startups teamed with Big Tech to gut the AI Act](https://corporateeurope.org/en/2024/03/trojan-horses-how-european-startups-teamed-big-tech-gut-ai-act) (March 2024) - CNN — [SpaceX acquires xAI](https://www.cnn.com/2026/02/02/tech/spacex-acquires-xai-elon-musk) (February 2026) - Washington Post — [Elon Musk's government contracts and funding](https://www.washingtonpost.com/technology/interactive/2025/elon-musk-business-government-contracts-funding/) (2025) - NPR — [Musk's DOGE group has access to federal payments system](https://www.npr.org/2025/02/04/nx-s1-5285403/musks-doge-group-has-access-to-the-federal-payments-system-what-does-that-mean) (February 2025)