Defense & Technology — 2026-05-09

White House Prepares Executive Order on AI Security Following Anthropic Mythos Model Cybersecurity Risks

The White House is preparing an executive order to create a vetting system for new AI models after Anthropic revealed its breakthrough Mythos model could find network vulnerabilities and pose a global cybersecurity risk. National Economic Council Director Kevin Hassett described an FDA-style approval process where AI models would be proven safe before release, while the Commerce Department expanded a voluntary testing program with Google, Microsoft, xAI, OpenAI, and Anthropic providing model access for government security assessment.

Analysis
Corroborated across Bloomberg and The Washington Post, the administration will likely sign an AI security executive order within 60 days of May 9. A signed order would establish the first federal pre-release vetting mechanism for high-capability models, shifting U.S. AI governance from voluntary disclosure toward structured regulatory gatekeeping. Hassett's public commitment to an FDA-style approval architecture imposes political cost on reversal. The Commerce Department's documented expansion of voluntary model-access testing with five major AI firms constitutes an observable institutional commitment that constrains quiet retreat. The voluntary testing program may instead be designed as a regulatory substitute, granting the administration oversight credibility without the binding pre-market approval obligations a signed order would impose.

2 sources

White House Studying AI Security Executive Order - Federal News Network
The secretive AI hacking tool fracturing Trump's White House - The Washington Post