Thank you.
Certainly, one of the key concerns of the EU AI Act is its implementation. It was first moving in the launch of a high-risk system, which is quite distinct from AIDA. I think it was something we wanted to see better implemented in Canada as a way of assessing and trying to understand where the applications of AI could be. Certainly, as an educator, the fact that AI in education was a high-risk application is a good example of how this act was developing a literacy to try to process and triage human rights risks for new AI technologies.
I think the concern has been the move towards generative AI models, which I think elude regulation strategically because they're meant to be all things at once. I think that is one of the things we need to be mindful of: both the way the EU AI Act has stalled around generative and foundational models and the way that the assessment and the techniques of assessment were largely delegated to third parties, meaning that there wasn't institutional capacity at the government or state levels to make these kinds of classifications.
I think those were two key gaps that we could learn from.
The third—to emphasize—is, in Europe, the development and proliferation of open models and the opportunity we have, in collaboration with other developed nations, to work towards more digital solidarities and stacks that allow us to take advantage of these complex technologies that one nation alone might not be able to create.