AI Use Policy

RevysAI is an artificial-intelligence system, not a person. No answer from it is written or reviewed by a human before it reaches you.

Answers are generated by third-party language models, reached by API from our server. The default provider is Anthropic, with the Claude Haiku 4.5 model.

If that provider is unavailable, the request may be served by an alternative provider configured for the service. The model currently in use is shown on the RevysAI screen itself — we would rather show you which one is running than always assert the same name.

Calls always originate on the server. The API key never reaches your browser.

Only what it needs to answer well:

• what you wrote, and the earlier messages in the same conversation;
• your first name, your school year and your language, so it can pitch the level and language of the answer;
• the selected subject, where there is one;
• any content you paste or ask it to summarise.

Not sent: your email, your password, your subscription details, your payment method, or the content of other conversations.

Only to produce the answer you asked for: explaining, summarising, creating flashcards, building questions, planning study or assembling a presentation. We do not use your conversations for advertising, profiling, or any analysis beyond keeping the service working.

The conversation is stored in your account, so you can find it again, until you delete it or delete the account.

At the AI provider, the text is processed to produce the answer and retained for a short period for security and abuse prevention, under their policy. Our providers do not use that content to train models.

Language models produce plausible text, not verified text. That means RevysAI can:

• state something false with complete confidence;
• invent dates, numbers, names, formulas and citations;
• confuse similar topics and mix up contexts;
• be out of date on recent facts;
• reproduce biases present in the data it was trained on.

Always check before using it in an exam, an assignment, or anything that carries a grade. Responsibility for what you do with AI output is yours — and the point of it here is to help you learn, not to study in your place.

Do not write to the AI: passwords, bank or card details, identity documents, health information, other people's data, or any secret whose disclosure would harm you.

If you send it anyway, that text passes through the AI provider like any other and stays stored in the conversation until you delete it.

Each plan has a monthly AI quota, and the size of the content sent per request is also limited. Quotas are enforced on the server, per account.

Circumventing the limits, automating calls, using the account as an intermediary for other people, or reselling access is prohibited. Accounts doing so are suspended.

We also block, as far as we can, requests seeking illegal or harmful content. The AI providers apply their own safety rules to the same request.

The EU AI Act requires that it be clear to a person that they are interacting with an AI system and that content is AI-generated. That is what this page and the notices inside the app do.

The use here — study support, with content the person reads and checks — decides nothing about anyone's life: it does not assess or rank students, and it does not determine access to education, employment or credit. It is therefore not a high-risk system under the AI Act's classification. If that changes, this section changes first.