President Trump, long champion of a hands‑off AI policy, is now weighing an executive order that would create a White House AI working group to vet new models before public release. Officials told executives from Anthropic, Google and OpenAI that the group could draft formal review procedures similar to Britain’s safety regime. The shift follows Anthropic’s debut of the powerful Mythos model.

Anthropic’s Mythos can spot software vulnerabilities so effectively that the startup refused to open it, warning of a potential cyber‑reckoning. Washington worries a breach could expose U.S. infrastructure and is also assessing whether such capabilities might serve the Pentagon or intelligence agencies. A lingering dispute over a $200 million defense contract has already halted Anthropic’s tools in military projects.

The proposed review would likely involve the NSA, the Office of the National Cyber Director and the Director of National Intelligence, filling a gap left by the sidelined Center for AI Standards and Innovation. Industry leaders warn that excessive oversight could slow U.S. innovation against China, yet the administration argues that pre‑release vetting protects national security while preserving market growth.