EC Midday press briefing of 01/09/2026
1 septembre 2026
EC Midday press briefing of 01/09/2026
Plus d'options
00:00 - 00:47
Intervenant 1
Good morning, Gian Volpicelli, Bloomberg News. I have a question regarding the AI Act requests for information that were recently sent to a certain number of AI labs. At the same time these RFIs were being issued, I was reading in horror the reports of several incidents regarding models developed by Anthropic and OpenAI, specifically in the latter case about 700 autonomous agents coordinated to hack a company. So I was wondering whether both Anthropic and OpenAI have duly received an RFI regarding the cyber risks of their models.
00:47 - 01:54
Intervenant 2
Hello Gian, and thanks for the question. So let me start by confirming what Executive Vice-President Henna Virkkunen has announced. So we have sent a number, and these are the first RFIs, so enforcement action has started on the Commission side against more than 30 AI companies in two main areas. The first one is safety and security, generally speaking, including to general-purpose and the most advanced AI models. And then we have a second set which is more on copyright and transparency issues that has been sent to another number of AI companies.
Now, we will not be naming these companies for a very clear reason, is that these are simple RFIs, these are the first enforcement steps, our dialogue continues with all companies. Now, on the two main companies that you're referring to, let me recall that we take cybersecurity risks of course extremely seriously. We have been very vocal on this over the last weeks and months, and our work with both Anthropic and OpenAI continue and we're having close contact also with them obviously.
01:54 - 02:36
Intervenant 1
Yeah, just quick follow-ups. The first one is whether in the wake of these incidents, which were vastly reported, widely reported, there's been any specific contact with these companies regarding specifically the incidents. And the second question, which is a bit of a nostalgia operation, but has ENISA finally secured access to the Mithos model? Because that was announced, there was an announcement of an invitation, from what I understand ENISA and the Commission were in talks with Anthropic to specify the terms and conditions, but I don't know of any updates regarding whether we actually have the beast.
La suite de la transcription est réservée aux clients Pro
Débloquez la transcription complète, le chat et toutes les fonctionnalités d'analyse avec Polyfact Pro.