The White House is considering pre-release reviews for high-risk AI models, including Anthropic's advanced vulnerability-discovering system, Mythos. This potential policy shift would mark a change from President Trump's earlier revocation of Biden's AI safety executive order, though no formal actions have been taken yet.
The deliberation underscores ongoing federal versus state regulatory debates and highlights the dual-use risks of advanced AI. Prediction markets currently price 100% YES for Mythos being provided to the US government by April 30, 2026, but increased regulatory scrutiny could alter that outlook.
Key actors to watch include Federal Chief Information Officer Gregory Barbaccia and President Trump. Any formal executive orders or OMB memos could impact Mythos's deployment, as could feedback from civilian agencies and the model's cybersecurity performance.