Hegseth Wants Pentagon to Drop Anthropic's Claude — But Military Users Are Resisting
Despite Pentagon leadership designating Anthropic a "supply-chain risk" following a dispute over AI safety guardrails, Pentagon staff, IT contractors, and former officials are resisting the order. They describe Claude as superior to available alternatives — with one contractor noting xAI's Grok "often produced inconsistent answers to the same query."
Defense Secretary Hegseth ordered a six-month phase-out starting March 3. Recertifying systems that run on Anthropic's products for military use is expected to take months. OpenAI and Google are reportedly negotiating to fill the gap on non-classified systems. Meanwhile, Anthropic's court filing revealed the DoD privately told the company sides were "nearly aligned" — just days before the public break.
The standoff is a test case for a fundamental question: can AI companies impose safety limits on government customers, and who wins when sovereign procurement power clashes with a lab's usage policies? The answer will shape how every major model vendor structures defense contracts going forward.