We, Europeans, produce tables of contents

There is a lot of talk lately about Article 50 of the AI Act and the Code of Practice on Transparency of AI-generated Content.

My impression is we have the old saying again: “America innovates, China copies, Europe regulates.”

Years ago when I was in Brussels for the review of a research project financed by the European Commission, I remember saying during a lunch how sad it is that we are so far behind America. By we I meant the IT sector in Europe.

A certain professor didn’t agree with me at all. He said we might not have Microsoft, Google, or Facebook, but we are far ahead in terms of web services.

I have no idea what he meant. Something interrupted our discussion and I never got a chance to ask him. But I guess he meant nice European companies that offer web services like Amazon, Cloudflare, or Stripe.

I am not the one who could complain about European research projects, I spent way too many years earning my money this way. But I can say people considered a very successful video call when we agreed on the table of contents for some bloody document we needed to write.

Of course, writing the bloody documents goes faster today (thanks to the AI tools produced by US companies). But it seems the general pattern is similar. We, Europeans, produce documents.

I have nothing against Article 50. In fact, I think it’s very much needed. But it won’t prevent situations like how a social media campaign triggered a migrant surge on Ceuta, or how misinformation lured thousands towards Europe. Around Ceuta, recycled clips, miscaptioned videos, and AI-generated images circulated alongside real footage, and nobody could tell which posts showed what actually happened. There are deeper actions needed.

Actions like building cryptographic proofs into the media pipeline. That means signing photos and videos at capture, and proving what was edited afterwards. Because adversarial accounts will not follow the Act and “no AI label” is not evidence that something is authentic.

Or actions like finally deploying digital identities and requiring them for posting on social media. That one might be controversial, but I am not saying that people should not be able to post anonymously. They should. And selective disclosure should be enabled: prove you are a real person (or that you are old enough, or that you are a resident of a particular country) without handing the platform your full identity. And if you are caught posting fake videos, the platform should still be able to ban you.

Only with measures like these will we be able to tell which images and videos were genuine recordings of what they claimed to show.