White House drafts AI pre-release safety testing order

The Trump White House is drafting an executive order requiring mandatory pre-release government safety testing for frontier AI models, confirmed by NEC director Kevin Hassett. The policy was directly triggered by Anthropic's Mythos model exposing measurable cyberattack vulnerabilities. AI labs deploying frontier models should prepare for federal evaluation checkpoints before public release, marking a structural regulatory shift from voluntary commitments like the 2023 Biden-era safety pledges toward enforceable pre-deployment review.