Abstract:
A federal appeals court in Washington, D.C., issued a ruling Friday upholding the Pentagon’s blacklist designation of Anthropic. AI company Anthropic has suffered a major setback in a months-long legal battle between the company and the Trump administration. The panel voted 2 to 1 to reject Anthropic's claim. Anthropic previously argued that the Department of Defense’s decision to ban Claude’s large model was an arbitrary administrative act, lacked statutory authority and was unconstitutional.
The majority opinion was written by Judge Gregory Catslas of the Federal Court of Appeals for the District of Columbia, joined by Judge Naomi Rao (both appointed by President Trump): "The Department of Defense has sufficient basis to conclude that continued access of Claude to Department of Defense information systems, whether by the military itself or by its contractors, would constitute a national security risk as defined by law."
Judge Karen LeCraft Henderson, appointed by President George H.W. Bush, issued the dissenting opinion.
In March of this year, negotiations between the two parties on how the military would use Claude's model completely broke down. The Department of Defense immediately designated Anthropic as a supply chain risk and determined that the company posed a threat to U.S. national security. The designation not only prohibits the use of the Anthropic model by the U.S. military, but also prohibits all defense contractors from using the model in Department of Defense projects.
Anthropic filed lawsuits in San Francisco and Washington, D.C., hoping to have the blacklist revoked. The Ministry of National Defense conducts risk determination based on two independent sets of identification documents, so it needs to respond to lawsuits in two different courts.
Last month, a federal judge in San Francisco ruled that one of the findings was illegal, and on Friday, the Washington, D.C., appeals court upheld the second finding.
Anthropic spokesperson issued a statement to : "We reserve our opinion on the court's decision. Another federal court has ruled that the government's parallel set of determinations is unlawful. We stand by our position and are considering all avenues for relief, including continuing to apply for review."
The appeals panel ruled that the stay of judgment is effective immediately, giving Anthropic time to apply for a retrial from the same panel or to the Court of Appeals for the District of Columbia Circuit for an en banc review; Anthropic can also submit the case to the U.S. Supreme Court.
Before the conflict became public this year, Anthropic was an early partner of multiple U.S. government agencies, including the Department of Defense. In July 2025, Anthropic signed a contract with the Pentagon. But in September of the same year, talks broke down between the two sides over deploying Claude to the Department of Defense AI platform genai.mil.
The Department of Defense requires Anthropic to open the model to the military to support all legal uses without any restrictions; and Anthropic wants to be guaranteed that its technology cannot be used for fully autonomous killing weapons, nor can it be used for domestic mass surveillance.
"The Department of Defense has reasonable grounds for concern that Anthropic may alter the underlying design of the Claude model and refuse to perform national security tasks that the Department of Defense believes are authorized and necessary under the contract," Judge Katzlas wrote in Friday's ruling.
Comments