Loading page…
The divided ruling keeps a military procurement ban in place despite a separate California decision against a broader restriction.
Listen to this story
By a 2–1 vote, a federal appeals panel rejected Anthropic’s legal challenge to the Pentagon’s March supply-chain-risk designation, leaving the ban on Claude in Defense Department systems and contractor use in force. The court accepted that a model whose provider can enforce limits—including against fully autonomous lethal weapons and mass domestic surveillance—could pose a covered integration risk; it did not find Claude caused a security incident.
Judges Gregory Katsas and Neomi Rao formed the majority; Judge Karen LeCraft Henderson dissented.
Anthropic says the blacklisting cost it billions in lost business, but has not provided a further update on its financial impact.
The majority rejected Anthropic’s due-process and free-speech claims, treating the exclusion as a response to its contract refusal.
Anthropic refused to lift limits on how the military could use Claude. Now a divided federal appeals court has upheld the Pentagon’s decision to keep the company out of its supply chain. Friday’s 2-1 ruling leaves the exclusion in force and gives the Pentagon a legal victory in a dispute over who controls the AI model’s use.
Circuit Judges Gregory Katsas and Neomi Rao rejected Anthropic’s argument that the Defense Department’s supply-chain risk designation was arbitrary, unauthorized and unconstitutional. Judge Karen LeCraft Henderson dissented. The majority focused on Claude’s continued use in department information systems, including systems used by contractors.
Anthropic can encode restrictions into Claude that stop tasks it does not want the model to perform, the majority noted. The judges accepted the department’s conclusion that continued integration of a model subject to those controls presented a risk covered by the supply-chain law. That is the court’s legal finding, not a finding that Claude had caused a security incident.
The conflict began in February when Anthropic refused a demand to remove restrictions on using Claude for fully autonomous lethal weapons and mass domestic surveillance of Americans. Defense Secretary Pete Hegseth regarded the company’s position as a significant national-security risk. The Pentagon made its supply-chain risk designation in March.
The designation canceled Anthropic’s military contracts and barred other Defense Department contractors from using its technology. Anthropic has said the blacklisting cost it billions of dollars in lost business and harmed its reputation. That loss figure is the company’s account; it has not provided a further update on the designations’ effect on its bottom line.
The majority also rejected Anthropic’s due-process and free-speech claims. It described the exclusion as the result of Anthropic’s refusal to accept a contract term the department considered essential, rather than punishment for the company’s support of tighter AI regulation. That distinction undercuts one of Anthropic’s challenges to this particular designation.
Friday’s decision does not erase Anthropic’s win in a separate California case. A federal judge there had thrown out another supply-chain risk designation and confirmed that decision last month. The two designations were challenged in different courts under separate laws. The Washington ruling leaves the Pentagon’s blocking of Anthropic in place despite the California result.
The Washington panel had already declined in April to pause the designation while the case proceeded, finding that Anthropic had not met the demanding standard for immediate relief. Friday’s ruling decides more than that temporary request: it rejects the company’s challenge to this designation. It does not make the separate California decision disappear.
Anthropic says it disagrees with the appeals court and is considering further review. Its options could include asking a larger panel of the Washington appeals court or the Supreme Court to hear the case. For now, the practical result is clear even if the wider litigation is not over: the Pentagon can continue excluding Claude under the designation upheld on Friday.
Editorial analysis
The key distinction is between winning a challenge to one government restriction and regaining access to Pentagon work. A California court rejected a separate designation, but Friday’s ruling leaves this exclusion standing. In our view, that gives the Pentagon leverage in disputes with AI suppliers over who sets limits on military use, without settling every legal question Anthropic has raised. The next concrete test is whether Anthropic seeks review from a larger appeals panel or the Supreme Court—and whether either court changes the status of the designation. Until then, the split outcomes are more important than either side’s claim to have won the wider dispute.
Loading discussion...
Join the conversation
Who should get the final say over those limits?
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.