When you take part in a session, you send text, perhaps audio, and finally a submission. Here you can see where it goes, who processes it, and what is left behind. The map first shows where the data is stored. Choose written or spoken dialogue to see how the conversation flows.
The AI colleagues reply using Mistral AI, a French company. We use their EU endpoint, under an agreement that the conversations are not stored.
In the spoken rooms, Speechmatics turns the audio into text. The audio is not stored by us.
In spoken rooms, the colleague’s reply goes to ElevenLabs in the USA, which reads it aloud. Never your voice, your name or what you said. Otherwise, something goes to the USA only if you choose to create your own avatar.
We use Anthropic’s Claude only in the course responsible’s tools. None of them reads the conversations, the submissions or the information about you.
Filled points are places where something about you is stored. The rings process data along the way, but store nothing for us. Select a place to see what is stored there.
You enter the code from the course responsible. You do not need an account.
Student number and name are asked for only when the session requires it, for example to tick off a compulsory activity. Otherwise you are “Participant 07” from the start.
In the 3D rooms we record how long you spend in each zone, how long your conversations are, and whether you switch tab. Never your position, and never what you wrote or said.
Answers you write or choose are stored in the database. Files go straight from the browser to a private file storage at Scaleway in Paris. They can only be retrieved through us, by you and by the course responsible for the session.
You can delete a file before you submit. After you have submitted, it is locked.
After a session, the course responsible may ask you to answer a survey. Opening the link registers nothing. The answer is stored only when you press “Start”, and you can choose “No thanks” instead.
Surveys are anonymous by default: the answer is stored without name and student number, and the course responsible does not see who answered. What is stored is which session and which condition (A or B) you were in, and a code that prevents you from answering twice. If the survey is not anonymous, the start page says so, and the answer is linked to your participation in the session.
Whether the answers can also be used in research is a separate choice on the start page, independent of the choice you made when you joined the session. Nothing is preselected, and the research export includes only answers where you have said yes.
If you choose to create your own avatar, your free-text description goes to Mistral, which turns it into a short list of details. Only that list goes on to Google (Gemini) in the USA, which creates the face image for the avatar. No image of you is sent.
The image is stored in the file storage and can be read by anyone who has its address. Your description is also stored. This is voluntary; you can choose a ready-made avatar instead.
The course responsible sees their session through us. By default you are shown as a pseudonym. Name and student number are shown only when the course responsible turns that on, for example when someone has typed the wrong number.
Transcripts for research include only those who have consented to research. Consent is voluntary, and you can take part in teaching sessions without it.
The notepad in the room is stored only in your browser and is never sent to us.
The website has no analytics or tracking scripts, and the fonts are loaded from our own server, not from Google.
All the services we send student data to, what they receive, and where.
| Service | Applies to | Used for | What they receive | Where | Storage |
|---|---|---|---|---|---|
| Upsun (Platform.sh) | All sessions | Operation of the website and the database | Everything stored about the session | France | For as long as the session exists (see below) |
| Scaleway | All sessions | File storage | Submitted files (private storage) and avatar images | Paris, France | For as long as the session exists |
| Mistral AI | All sessions | The AI colleagues, written and spoken | What you write and say (as text), and your nickname | EU endpoint | Agreement of zero storage; not used for training |
| Speechmatics | Spoken rooms only | Speech to text | Audio from the microphone while you hold the key, and your IP address | EU endpoint; operational logs can be read from the USA | Audio is not stored; the IP address stays in operational logs for about a month |
| ElevenLabs | Spoken rooms only | Reading the colleagues’ replies aloud | Only the colleague’s reply | USA | The audio is not stored by us |
| Google (Gemini) | Only if you create your own avatar | The image for “Create your avatar” (voluntary) | A short list of appearance details | USA | The image is stored by us |
We do not yet have a fixed retention period for data from the sessions. What is stored from a session stays until it is deleted, and today we do the deleting by hand. A fixed retention period with automatic deletion is being worked on, and this page will be updated when it is in place.
If you want to know what is stored about you, or have it deleted, write to kontakt@virtualaicorp.no.
Course responsibles use more tools than students, including Claude from Anthropic for building simulations. None of these tools is in contact with data from the students.
How we handle data for course responsibles →
Last updated 6 October 2026.