Google, Microsoft, and xAI allow the government to test AI models in advance

Google, Microsoft, and xAI will let U.S. government evaluators test unreleased AI models before public release.

The Center for AI Standards and Innovation will examine risks tied to cybersecurity, biosecurity, and potential misuse.

The effort expands earlier industry-government testing and signals a more active federal role in AI governance.

The move could shape how frontier models are reviewed for security and broader public-risk concerns.