AI information

1. Provider

2. What the AI is used for

• Answers in the chat: The AI writes answers from the knowledge entries the business has entered in the dashboard.
• Reading a website: At the business’s request, Hevia proposes entries from the text of its website. The business checks every proposal itself.

The AI makes no decisions about people and no automated decisions within the meaning of Art. 22 GDPR.

3. What the AI receives

The question, the last messages of the conversation, the matching knowledge entries and the name, type and town of the business. Not sent to the AI: IP address, details from forms in the chat, and data from other conversations or businesses.

4. No training, storage at the provider

Under its Commercial Terms, Mistral does not use the data to train models; Hevia does not train any models on conversations either. Mistral stores requests and answers for as long as needed to produce the answer, plus up to 30 days to detect abuse.

5. Limits

• The AI can make mistakes. Binding information, prices, bookings and appointments are always confirmed by the business.
• If the entries contain no answer, the AI is meant to say so and offer contact with the team instead of guessing.
• If the AI is unavailable, Hevia answers directly from the entries without AI.
• Special categories of personal data (such as health data) do not belong in the chat; the chat points this out before it starts.

6. Labelling as AI

Under Art. 50(1) of the AI Act, Regulation (EU) 2024/1689, people must be able to tell that they are interacting with an AI. Hevia therefore shows a fixed notice at the start of every conversation and an AI badge next to the assistant’s name. The business cannot switch either off.