Appeals court upholds the Pentagon's exclusion of Anthropic under a second statute

A panel of the US Court of Appeals for the District of Columbia Circuit upheld, by two votes to one, the Department of War’s order excluding Anthropic and its Claude models from department systems and contractor work. The Pentagon had designated the company a supply-chain risk in March after Anthropic refused to drop its restrictions on fully autonomous weapons and domestic mass surveillance. It relied on two separate statutes, so the fight was litigated in two courts. In August a San Francisco district judge struck down the designation made under 10 U.S.C. § 3252, whose definition of supply-chain risk centres on an “adversary”. The D.C. Circuit’s ruling concerned the other designation, an order under the Federal Acquisition Supply Chain Security Act (FASCSA). That law defines the risk more broadly, as one that “any person” might sabotage or manipulate a system.

Judge Gregory Katsas, joined by Judge Neomi Rao, held that the department “had ample support for its conclusion that the continued integration of Claude into the Department’s information systems, by the Department or its contractors, presented a statutorily covered national-security risk”, CNBC reported. The majority accepted that restrictions trained into a model could constrain military operations even though Anthropic could not alter a model after deployment. It noted Secretary Pete Hegseth’s concern that “overly constrained” models might stop working unexpectedly. According to Holland & Knight’s summary, it also found that the order was prompted by Anthropic’s refusal of a specific contract term, not by its public advocacy on AI safety. It treated the department’s failure to follow the statute’s notice procedures as harmless error. “In our Republic, it is the President and the Secretary of War who must determine how best to balance the competing risks,” Katsas wrote.

Judge Karen LeCraft Henderson dissented. According to Clark Hill, she read the statute as aimed at deliberately subversive conduct by hostile actors, not a vendor’s openly disclosed usage restrictions. She warned that the majority’s reading would let the government treat almost any vendor-imposed limit as a national-security risk.

The court stayed its judgment for seven days so that Anthropic could seek rehearing by the full circuit. “We respectfully disagree with the court’s decision,” an Anthropic spokesperson said, adding that the company was “considering all options, including further review.” The district court’s injunction against the § 3252 designation still stands, so the outcome is split. The department can continue to direct contractors to remove Anthropic products under FASCSA, ahead of its 30 September deadline for ending all Anthropic-related work.