[UPDATE] White House Weighs Mandatory AI Pre-Release Reviews

Less than a week after Google, Microsoft and xAI agreed to voluntary pre-release testing with the US Center for AI Standards and Innovation, the Trump administration is now considering an executive order that would make such reviews mandatory for the most capable models.

The Washington Post first reported on May 8 that internal discussions are underway around a working group of tech executives and government officials, tasked with examining what a formal review process might look like. Kevin Hassett, director of the National Economic Council, has compared the proposed approach to the way the Food and Drug Administration evaluates drugs before they reach patients.

The catalyst is Anthropic's Claude Mythos Preview, the cybersecurity-capable model that the company has so far declined to release publicly. Mythos can autonomously identify and exploit zero-day vulnerabilities in major operating systems and browsers, and has reportedly found thousands of high-severity flaws during testing. Anthropic announced its Project Glasswing initiative to deploy the model for defensive cybersecurity work, but kept it out of general availability. Officials told the Post that the prospect of a similar capability falling into adversary hands shifted the calculus inside the West Wing.

The shift is striking. President Trump's earlier AI policy, written largely under former AI czar David Sacks, leaned hard against pre-deployment regulation. Sacks departed in March. White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent have since told associates they intend to take a more active role in shaping the administration's AI agenda.

The proposed executive order is still under discussion and has not been signed. Tech industry reaction has been wary, with several lobbying groups warning that mandatory pre-release vetting could slow product cycles and put US developers at a disadvantage relative to Chinese rivals. The Hill reports that early drafts of the order would task the National Institute of Standards and Technology and the Commerce Department's CAISI with drawing up the review framework.

The Mythos question runs in parallel to the recent voluntary CAISI agreements and the Five Eyes guidance on agentic AI. Any mandatory review regime, if signed into law, would represent the largest expansion of federal AI oversight to date.

Further reading: Washington Post, Fortune, Federal News Network, and The Register.