The conflict began in February, when President Donald Trump and Defense Secretary Pete Hegseth accused Anthropic of endangering national security and designated the company a supply-chain risk. The dispute centered on the Pentagon’s demand to use Claude for “all lawful uses,” against Anthropic chief executive Dario Amodei’s insistence on guardrails barring fully autonomous weapons and mass domestic surveillance.¹

Anthropic’s position was that those limits were not a refusal to support government work, but a line against uses it considered dangerous. The company declined to retreat as officials pressed for broader access, setting up a fight over whether a contractor can restrict how the military deploys its AI.²

On Friday, the U.S. Court of Appeals for the District of Columbia Circuit ruled 2-1 for the Pentagon, rejecting Anthropic’s challenge and allowing the department to keep Claude out of its systems and bar the company from Defense Department work. Judge Gregory G. Katsas wrote that the department “reasonably feared” Anthropic could alter Claude to block national-security functions it viewed as authorized and necessary.²

That ruling embraces the government’s central argument: that the risk lies not only in what an AI system does, but in whether its maker can constrain missions the Pentagon says it needs to carry out. Anthropic, however, said it “respectfully disagrees,” pointing to a separate California ruling this summer that found a parallel government designation unlawful. It said it was weighing further review.²

The decisions leave the company with a split legal landscape—and a sharper political question: whether AI safety guardrails are prudent limits, or an unacceptable constraint on military power.