The White House has a framework for how to review advanced AI models, according to an administration official, but did not detail it or say whether it would be released to the public, reports Politico
- AI companies are expected to review a draft Tuesday at a meeting with the Office of the National Cyber Director, according to four people familiar with the planning who were granted anonymity to discuss internal deliberations.
- “The voluntary framework outlined in the June 2nd executive order was complete by the deadline,” said a White House official, granted anonymity to discuss internal policy deliberations. “Discussions with industry about next steps are underway.”
- Meta (META), Anthropic and Google (GOOGL) will be among the companies in attendance at the ONCD meeting, three of the people said.
Frameworks of this kind have historically followed a familiar sequence: executive order, voluntary industry commitments, then the long argument over whether voluntary becomes mandatory. The operative word here is voluntary, which is what separates this from binding regimes of the sort the EU has legislated; voluntary US frameworks have tended to function as a floor that large labs accept readily and a signal of where eventual statutory or agency-level requirements would land. The transmission to the named megacaps has generally been muted at the headline stage, since the cost of compliance sits against capex already committed to frontier model development, and the material risk has instead been the follow-on: export control tightening, procurement conditions, or disclosure obligations that touch the training and deployment pipeline. What is worth watching is whether the framework is published and whether it converges with the safety-evaluation practices the labs have already announced, since convergence makes it a nonevent and divergence makes it a negotiation. Attendance lists of this kind, the major labs at the table, have in past episodes indicated the industry is shaping the text rather than contesting it.