White House AI Framework: Open Models Excluded, Access Restricted in Pre-Release Review
The White House is drafting an AI framework that labels 'covered frontier models' as closed-source with potential national-security risks, explicitly excluding open models while stating nothing should prevent their release. During a 30-day pre-release review, access to these models would be tightly controlled and logged in high-security environments, with a multi-agency final review and no mandate for public release. The process has included staff-level industry briefings, leaving many companies outside the loop, and there is no clear definition for what counts as 'state-of-the-art' or a risk, potentially sidelining less-advanced firms.