US government now has pre-release access to AI models from five major labs for national security testing

The Commerce Department’s CAISI now has pre-release access to models from five major AI labs for national security testing.

New agreements with Google DeepMind, Microsoft, and xAI expand earlier deals with Anthropic and OpenAI.

The agency can evaluate advanced, sometimes de-safeguarded models in classified settings to look for security and cyber risks.

The move gives the U.S. earlier visibility into frontier AI capabilities as competition with China intensifies.