Anthropic is taking its fight with the Pentagon to the full bench of a federal appeals court. On Tuesday, October 13, the AI company filed a petition asking every active judge on the federal appeals court in Washington, the D.C. Circuit, to rehear its challenge to the Defense Department's decision to label it a national-security supply-chain risk, according to Reuters, CNBC and Bloomberg Law.
The petition targets a September 25 ruling in which a three-judge panel sided with the government by a 2-1 vote. Court records list the opinion as written by Judge Gregory Katsas, with Judge Karen LeCraft Henderson in dissent, in two consolidated cases, Nos. 26-1049 and 26-1162. Reuters reported that the majority found the Pentagon had acted reasonably and accepted its view that Anthropic's safety restrictions could compromise military operations, while rejecting the company's claim that it was punished for its views on AI safety.
The dispute dates to March, when the department, named in court papers as the Department of War, applied the label after negotiations over how the military could use Anthropic's Claude models broke down. CNBC reported that, as a result, defense contractors could no longer rely on Anthropic's technology when doing work for the department. Anthropic and the Pentagon had agreed a $200 million contract in July 2025, and talks collapsed later that year as the parties negotiated Claude's deployment on the GenAI.mil platform. At the heart of the disagreement was Anthropic's refusal to let its models power mass surveillance of U.S. citizens or lethal autonomous weapons.
In the new filing, Anthropic's lawyers argue that the panel majority misread the procurement statute, turning a provision meant to address hostile foreign actors into one that could presumptively reach any government contractor. They also contend the court allowed the Pentagon to act without any process before the designation and despite what the company calls clear evidence of retaliation for its public statements on AI safety. The company has previously put the commercial damage from the blacklisting in the billions of dollars.
The appellate fight runs alongside a separate case that has gone Anthropic's way. In August, a federal judge in San Francisco ruled that the department had violated the company's First Amendment rights and struck down a parallel designation issued under a different law, both CNBC and Reuters noted. The two tracks leave Anthropic with conflicting outcomes from different courts over closely related government actions.
The stakes extend beyond one vendor. The case tests how far national-security procurement powers can be used against a domestic AI supplier whose usage policies conflict with military demands, a question that matters to every lab that sets limits on how its models may be deployed. Dario Amodei, Anthropic's chief executive, has urged rules that would apply to every U.S. frontier AI developer, Reuters noted, while the administration has favored a lighter, pro-growth approach.
A petition is only a request. The full court can decline to rehear the case, and the panel ruling stays in force unless it is vacated. Anthropic declined to comment beyond the filing, and a Pentagon spokesperson told CNBC that the department does not comment on pending litigation. The petition text itself was not available for review, so its arguments are described here as reported by the three outlets.