White House exempts open AI models from security review

The White House is reportedly excluding open AI systems from a voluntary security review framework for advanced models, drawing an important line between closed frontier systems and open-weight releases.

Washington Post, Politico, Axios and other outlets report that administration officials briefed major technology companies on a framework that would apply to the most advanced closed models while leaving certain open systems outside the review process. Axios says the framework is not being made public for now.

That distinction matters because open-weight models are central to the current policy fight. Supporters argue that open releases help researchers, startups and national competitiveness. Critics worry that powerful openly available models can be adapted for cyber, biological or other misuse with less oversight.

The reported framework does not appear to ban model releases. It is a voluntary review process, but it signals where Washington wants scrutiny to land first: the largest closed AI systems from major labs, rather than broadly available open models.