U.S. Will Now Examine National Security Implications of New AI Models, Pre-Release

The U.S. government has established parallel tracks to assess and use frontier AI: CAISI (under Commerce/NIST) will evaluate companies' frontier models for national-security risks before public release, while the Department of Defense will deploy selected providers' models on classified Impact Level 6–7 networks for operational support. Agreements cover major vendors (Google DeepMind, Microsoft, xAI, OpenAI, NVIDIA, SpaceX, AWS, Oracle, etc.), and Anthropic’s exclusion from Pentagon deployments—despite its Mythos model drawing concern—highlights a policy tension between risk assessment and operational adoption.