Regulation2026-08-03
The White House is meeting representatives of OpenAI, Anthropic, Google and Meta to discuss a framework that would let the US government review the most advanced AI models before public release. The scheme comes from an executive order signed in June and participation by companies is voluntary. Most of its standards will be classified, and it is still unclear who in the administration will run the review or whether freely downloadable models are covered.
What changed
US oversight of advanced AI models rested on company self-policing, with no government pre-release review process.
What it unlocks
A route for the US government to examine the most advanced AI models before they are made public, on a voluntary basis.
- government access up to 30 days before public release
- EU AI Act in effect August 2
- fines up to 7% of annual gross revenue
- more than 1,200 staffers signed open letter
What you need to act on it
- voluntary company participation
- framework details largely classified
- edition.cnn.com2026-08-03