EU Opens Formal Talks With OpenAI and Anthropic After 'Agent Containment Failures'

New AI Act enforcement powers take effect, giving Brussels the authority to demand pre-release evaluations — and it is reportedly already invoking them following recent agent safety incidents.

The European Union has activated new enforcement powers under the AI Act, and it wasted no time putting them to use. According to @Nefarious_Trey, Brussels can now investigate frontier developers and demand pre-release evaluations, and has opened formal discussions with both OpenAI and Anthropic following what the report describes as recent agent containment failures.

Unlock the full briefing

Get every story in today's briefing, the full archive, and the daily AI intelligence brief.

All stories today

Full archive

Daily brief

Cancel anytime. Payments powered by Stripe.