Technical group

1. The standards and rules set by Gaia-X imply that each participant (such as a service provider) must be identified (by a self-description) and “labeled” Gaia-x (labeling is no longer necessarily compulsory for after Pierre GRONLIER, will be confirmed in January).

2. Cozy exists as an independent company, it also provides its white label service to third parties. Cozy is not about designing a data warehouse. Each user has their space in the form of a silo. The common computation for AI is organized by nodes with an encrypted random distribution in which the various contributors participate. No one can pull up the device's "calculation plan" and relate the data to its owners. The companies or institutions that Cozy provides act as data controllers (and Cozy is a processor)

3. How do I interface Cozy with the distributed identity? Today the user can connect to Cozy via his ENP, Cozy can provide or delegate authentication (for the Rennes experiment, EduConnect is used).At the moment there is no contracting layer in Cozy's AI. The link between the supplier and the user is hard-coded, so not industrialized. With “DASES / PROMETHEUS-X”, it will be different because it will require contractualization and consent. So Prometheus-X could help scale it up.
To generalize all the services on Prometheus-X, this identification / labeling / contractualization must be automated / industrializable.
The algo may be published and available to any user. Prometheus-X will need to provide this service through a micro contract to allow a user to use Cozy's AI. How to use Cozy's security solution (node ​​/ random / encryption) with distributed identity and enable AI training. How to identify the different parts and find out how the existing protocol can interface with these identities

4. Collaboration with MAIF and BNP can be interesting, such as setting up workshops, and using Prometheus-X as a portal by presenting the offer it provides

5. It is necessary to be able to change data host very simply at any time without friction

6. The data that will pass will be aggregated data but not anonymized. Cozy allows you to link several data sources (nest one service in another, which is nested in another and so on). UX big topic on how to run one service that appeals to another, and what are the responsibilities of each "service", who benefits from the end result?

From Matthias:

• Consent management for personal data from and to peri / para-school actors, for example orientation
o It is a question of differentiating between on the one hand the AI service on distributed training and on the other hand the consent for the exchange of personal data
o The AI service can be made available from the Prometheus platform and interoperated with other services so that the community can use it for various purposes
o Consent management can be integrated into the ENPE for various use cases, for example as part of the orientation
o The two can sometimes be linked (I give my consent to contribute to a calculation) but not always (I give my consent to send my data to an extracurricular actor so that he can have a complete profile of the person in his assistance
• We need to deal with both subjects with different use cases