Commerce Pulls AI Testing Site as Trump Team Fights Over Spy Role in Evaluating Models
Updated
Updated · The Washington Post · May 11
Commerce Pulls AI Testing Site as Trump Team Fights Over Spy Role in Evaluating Models
6 articles · Updated · The Washington Post · May 11
Commerce’s Center for AI Standards and Innovation took down a website announcing testing partnerships with Microsoft, Google and xAI, reflecting White House sensitivity as officials debate who should evaluate frontier AI models.
The internal fight centers on a proposal from the Office of the National Cyber Director to build a large evaluation center inside ODNI, which would give intelligence agencies a much bigger role than Commerce’s current voluntary testing system.
Mythos sharpened that debate after Anthropic said the model can find software vulnerabilities better than humans, raising fears it could help inexperienced hackers hit grids, banks and government agencies.
Trump could sign an AI security executive order as soon as Monday, but officials remain split over mandatory versus voluntary testing and over how far Washington should go without undercutting U.S. firms competing with China.
Can an 'FDA for AI' stop cyber threats without crippling America's tech race against global rivals?
AI can now autonomously hunt for software flaws. Who is building the AI to defend against it?