Appeals court lets Pentagon keep Anthropic blacklist

A US appeals court said the Pentagon can keep Anthropic on a supply-chain risk blacklist, reviving a major legal threat to Claude’s role in federal and defense work. The ruling says the Defense Department had authority under a broader procurement law to restrict Anthropic, even without showing that the company acted with malicious intent.

The decision complicates an earlier lower-court ruling that had blocked a parallel designation. That judge said Anthropic did not fit the narrower definition of a supply-chain risk because the government had not shown sabotage, hidden access or another adversarial technical threat. The appeals court said a different statute gives the Pentagon more room to act when it sees operational risk.

The dispute grew out of Anthropic’s limits on military uses of Claude, including lethal autonomous weapons and mass surveillance. Ars Technica reports that the appellate ruling framed the case as a hard tradeoff: constrained AI systems could fail during military operations, while unconstrained systems could hallucinate dangerous targets.

For AI companies, the ruling is a warning that safety policies are becoming procurement risk, not just product policy. Labs selling to government customers may need to choose between strict use limits and access to defense contracts, while agencies gain more leverage to demand models that behave the way military workflows require. Anthropic says it disagrees with the decision and is considering further review.