In short
Face Up lets trainers build role-play situations where the learner talks to an interlocutor driven by artificial intelligence, then receives a debrief. You always know when you are talking to an AI. Anything that came out of a model is marked as such. The debrief is for training purposes and cannot support any decision about your job, your pay or your career. Your data is hosted in France and is not used to train the models.
1. What Face Up does
Face Up lets a trainer build conversational simulations so their learners can practise the skills that matter when dealing with people.
- During the session: the learner speaks with a persona driven by artificial intelligence, playing the professional role they actually hold.
- After the session: a debrief is generated from what was said, to surface what to work on next.
- For the trainer: live observation, session replay and the debrief are the raw material for the teaching work.
2. Where we stand under the AI Act
European regulation 2024/1689, known as the AI Act, governs how artificial intelligence systems are placed on the market in the European Union.
- Our role: QONVERS., the company behind Face Up, is a provider of an AI system. The client deploying the solution to its teams is a deployer.
- Our category: Face Up does not fall under any prohibited practice listed in article 5, and is not a high-risk AI system within the meaning of annex III. It sits in the minimal-risk category, subject to the transparency obligations of article 50.
- Our analysis: this assessment was established with the support of an independent law firm.
3. You always know when you are talking to an AI
In line with article 50(1) of the AI Act, you are told before the first exchange, and the notice stays visible for the whole session.
- The notice “Interaction with an AI” appears on the screen before the session and on the session screen itself.
- It is visible without looking for it, in video and in audio, in fullscreen and in reduced display.
- It cannot be dismissed. There is no setting to remove it.
- Nothing in the product tries to make you believe you are talking to a real person.
4. Anything that came from an AI is marked
Article 50(2) requires content produced by artificial intelligence to be identifiable. We draw three distinctions, because not everything you see in Face Up came out of a model.
- “Interaction with an AI”, on the video and audio replay of your session and on the transcript. That content contains you: your own words and your own image sit next to the persona's. Presenting it as AI-generated would be false.
- “Content generated by AI”, on a scenario introduction video and on the solution audio of a debrief. There, the image, the voice and the text come entirely from a model.
- The debrief notice, quoted below, on hot and cold debriefs.
These notices sit above the content they qualify, stay on screen for the whole playback of a video or an audio file, and cannot be dismissed. They are carried over to exported files and to content pushed to your learning system, where they come with a machine-readable marking.
5. What the debrief is for
The debrief is produced by artificial intelligence from what was said during your session. It carries an assessment of the situation you have just played, and your trainer has access to it.
Its purpose is educational, and the notice shown on every debrief says so in the same terms as our contractual commitments:
Debrief generated by AI, for training purposes. It cannot support any decision about the learner's job, pay or career.
- That notice is the same for the learner and for the trainer. We do not run two different stories depending on who is looking.
- The assessment is about a situation played in a practice setting, not about the person.
- Face Up is not a recruitment or selection tool.
6. Your data is not used to train the models
- What is said during your sessions is never used to train, fine-tune or improve artificial intelligence models, whether ours or our partners'.
- Data and infrastructure are hosted in France.
- Data is encrypted at rest and in transit.
- Audio and video recording is a workspace setting, off by default, switched on where relevant by your organisation's training team.
7. Human oversight
The trainer stays at the centre. They build the situations, choose what is observed, access the sessions and run the teaching debrief. Artificial intelligence produces material, it does not replace the trainer's judgement.
8. Your rights
Under the GDPR you have the right to access, rectify, erase, object to and port your data.
- Where Face Up is deployed by your employer or your training provider, they are the controller and QONVERS. acts as processor. Send your requests to them first.
- You can also contact our Data Protection Officer at dpo@face-up.fr.
9. Questions
For anything concerning the use of artificial intelligence in Face Up, write to us at contact@face-up.fr.
