Google, Microsoft, and xAI Agree to Share Unreleased AI Models with US Government for Safety Testing
Wall Street Journal·high signal
NIST announced that Google, Microsoft, and xAI will provide unreleased AI models to the Center for AI Standards and Innovation for pre-launch evaluation of national security and public safety risks, focusing on cybersecurity, biosecurity, and chemical weapons. The move was prompted by concerns raised after Anthropic's Mythos model launch last month. Notably, Anthropic is absent from the agreement following its public dispute with the Trump administration over AI ethics in military applications.