Why the public debate around high-risk AI safety is ignoring structural privacy

Suddenly general audiences in the EU are hearing about AI legislation in the news, and high-risk systems present a risk to… privacy (really, that’s the only angle being put forward in some news outlets). As if the GDPR didn’t contain anything that can be (and is) used to manage the data protection aspects of AI systems, and as if that was the sole risk assessed.

Once again: AI is another topic, like a Venn diagram with two large circles that overlap each other in part.

And while the vote of the AIAct in the European Parliament is an important step, presenting it as adoption of legislation – full stop – minimises the impact that the coming negotiations between the Commission, Parliament and Council can have on the final text.

I know that making legal concepts and legislative processes understandable to “Joe Public” is not easy and is sometimes a feat in and of itself. But if you are taking too many shortcuts, the message is no longer simplified but inaccurate – and maybe an AI system will then do a better job! (It won’t be a high-risk AI system, mind you.)

So if you’d like to have a clearer picture about what this WIP legislation means for you or if you have broader questions about AI governance, do reach out!