Draft framework sent to OpenAI, Anthropic and Google would let the government assess advanced AI models before release, as Sam Altman meets White House and lawmakers amid debate over safety and open models.
The Trump administration is nearing completion of a voluntary framework under President Donald Trump's June AI executive order that would ask companies to provide their most advanced models to the government for capability assessments before full public release, with an Aug. 1 deadline for agencies to finalize the approach. The Information reported that a draft was sent about two weeks ago to OpenAI, Anthropic and Google, while OpenAI CEO Sam Altman said he has reviewed the proposal and is in Washington for meetings with White House chief of staff Susie Wiles, lawmakers and others to discuss upcoming models, cybersecurity and U.S. competitiveness. The effort has intensified debate over how to define frontier models, whether open-source or open-weights systems should be treated differently from closed models, and whether voluntary safety testing could expand into broader government control over AI.