AI declaration
Our statement on the use of AI in Bep, under the EU Artificial Intelligence Act (Regulation (EU) 2024/1689). Version 1.0, last updated 25 September 2026.
Beta: use at your own risk
Bep is in beta. The models and services we use may still change. When they do, we update this page.
Bep is an AI system
Bep is provided by [NAAM]. In Bep you practise legal conversations with a client played by AI, and afterwards you get feedback from an AI coach. So in Bep you are always talking to an AI system, not to a person. This page explains where AI is involved, what Bep is intended for and where its limits are.
What the AI generates
| Part | Purpose | What is AI-generated |
|---|---|---|
| AI actor | Plays the client in a practice conversation, by text chat or voice | Every reply from the client, as text or as a synthesised voice. The actor can also end the conversation itself, just like a real client. |
| AI coach | Gives feedback after the conversation | The summary, observations, development tips and the verdict per criterion in the feedback report. |
| Scenario Builder | Helps teachers and trainers set up a practice scenario | A proposal for the client profile, the actor's instructions and the feedback criteria. The teacher reviews and edits it. |
Not generated by AI: the scenarios and assessment criteria as the teacher sets them, and of course everything you say or type yourself.
What Bep is intended for
Bep is intended for practice: law students, (junior) lawyers and other professionals use it to train their conversation skills in a safe environment. The feedback is there to learn from.
Bep is not intended for assessment. Do not use the feedback for grades, exams, admission, progression or other decisions about students or employees. We also do not store conversations or feedback, and teachers cannot see students' results in Bep.
People stay in charge
- AI can make mistakes. The feedback may be incomplete or wrong, including on the law. Treat it as a tool to reflect on your conversation and ask your teacher when in doubt.
- The teacher sets the frame. The scenario and the criteria the AI coach looks at are written or approved by a teacher or trainer.
- Not legal advice. The cases are fictional and nothing in Bep is legal advice.
- Do not use real data. Practise with the fictional case and do not share real or sensitive personal data.
Which models we use
| Use | Model | Where |
|---|---|---|
| Text conversations | Google Gemini 2.5 Flash, via Vertex AI | EU (Netherlands) |
| Feedback and Scenario Builder | Google Gemini 2.5 Pro, via Vertex AI | EU (Netherlands) |
| Voice conversations | Google Gemini Live | US |
| Voice conversations (only if an organisation has switched this on) | OpenAI GPT-Live | US |
The AI coach works with the text of the conversation (the transcript), not with your voice. Bep does not infer emotions from your voice or face and does not use biometric data. How we handle your data is explained in the Trust Center and the privacy statement.
Questions
Questions about AI in Bep, or spotted something that is not right? Email us at info@bep.nl. Are you an AI agent, or building one? See For AI agents.
Questions? Send an email to info@bep.nl