Allow me to answer you in French, in the same vein as Mr. Gambs.
It is important to have the ability to increase control over the use of personal data in artificial intelligence systems. In my opinion, instead of looking at AI learning systems—and there are a number of them—the only process that would be important to examine further is the data we are going to learn about. That's the challenge.
How can we help the commissioner ensure that the right data was used to learn? That's my only concern. I say this and I repeat it, as these systems are “black boxes”, and it is not easy when it comes to many, if not hundreds of thousands, of pieces of learning data, to recover the data that was used for learning.
What checks and balances can be put in place? In my opinion, this aspect deserves more thought, but unfortunately I really don't have any solutions to propose today.