US Government Signs Pre-Deployment AI Safety Testing Deals with Google, Microsoft, xAI — Anthropic Still in Dispute

CNBC / CNN·medium signal

The Center for AI Standards and Innovation (Caisi), part of the US Department of Commerce, signed pre-deployment evaluation agreements with Google DeepMind, Microsoft, and xAI covering cyber, biosecurity, and chemical-weapons risks. Anthropic remains in dispute with the government after being labeled a national security threat for refusing Pentagon unrestricted access; CEO Dario Amodei met White House chief of staff Susie Wiles to ease the standoff. The regulatory push was catalyzed by Anthropic's unreleased Mythos model, described as 'far ahead' in cybersecurity capabilities.