White House orders Anthropic to halt Mythos expansion; considers formal AI model vetting regime before release

The White House refused Anthropic's request to expand access to Mythos (a model) as part of Project Glasswing, and is now considering an executive order creating an 'AI working group' of tech executives and government officials to examine potential procedures, up to and including a government review process for model releases.

The shift represents a prior restraint regime where parties wishing to release highly capable new models would need to request permission—a departure from previous U.S. Frontier AI policy rhetoric and an informal, ad-hoc approach rather than formalized procedures.

Critics including Neil Chilson and Dean Ball warn that arbitrary and informal government decision-making on which models can be deployed to which parties favors the connected, prevents planning, and enables corruption; the precedent also reflects the European Union pressing Anthropic for access while the White House refuses to allow it.