A federal appeals court on Friday upheld the Pentagon's classification of Anthropic's Claude as a supply-chain risk, rejecting the AI company's bid to overturn the ban in a 2-1 decision. The ruling means the Defense Department can continue excluding Claude from its systems, while a separate designation against Anthropic remains tied up in litigation after a San Francisco court struck it down earlier this year. The dissent, by Judge Karen LeCraft Henderson, argued the statute should be read narrowly to cover only subversive manipulation, not a contract dispute.
The case stems from a $200 million Pentagon contract signed in July 2025 and the collapse of negotiations over terms. Anthropic refused to make Claude available for fully autonomous weapons or domestic mass surveillance; Defense Secretary Pete Hegseth called that an attempt to override military operational decisions. The majority, citing the possibility that restricted models could fail at critical moments, said the executive branch must balance those risks.
Business consequences are mounting. Anthropic said the designations cost it revenue as customers grew wary, and the company is still hoping to go public later this year. The Pentagon has not detailed how it is replacing Claude, though rival models from SpaceX, Google, and OpenAI are options—deals that have sparked internal objections at some of those firms. Anthropic says it is considering an appeal to the full D.C. Circuit or the Supreme Court.