Anthropic's Pentagon Ban Survives Appeal—For Now
A 2-1 DC Circuit ruling upholds the DoD's supply-chain designation of Claude. Legal experts say the fight has barely started.
By The Defense Desk
A three-judge panel of the US Court of Appeals for the District of Columbia has upheld the Pentagon's designation of Anthropic's products as a "supply chain risk" to national security—a ruling that allows the Department of Defense to bar use of Claude not just by its own personnel, but by private-sector employees working on defense contracts. [1]
The 2-1 decision, handed down on September 25, landed in Anthropic's disfavor in part because of panel composition. Two of the three judges—Gregory Katsas and Naomi Rao—are Trump appointees. Judge Gregory Katsas wrote for himself and Judge Naomi Rao 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." [1]
The majority's reasoning zeroed in on Claude's built-in safety guardrails. "As Anthropic admits, the company encodes restrictions into Claude that prevent the model from performing tasks that Anthropic wishes to prevent. On more than one occasion, these restrictions have stopped Claude from performing tasks requested by government users," Katsas wrote. [1]
Judge Karen Henderson, a George H.W. Bush appointee, dissented sharply. The Federal Acquisition Supply Chain Security Act of 2018—the statute used to designate Anthropic a supply-chain risk—was, she wrote, explicitly passed to protect against sabotage by malignant foreign powers, not an American company openly building safety and ethics guardrails into its products. [1]
Pentagon CTO Emil Michael, who has led the charge against Anthropic, wasted no time on social media: "the hammer of justice has smashed AnthropicAI['s] arguments. They are a Supply Chain Risk to the defense industrial base [and] warfighters will sleep better knowing that no private company will insert their opinions in the chain of command." [1]
Anthropic kept its response measured. "We respectfully disagree with the court's decision. Another federal court has already held the government's parallel designation unlawful," the company said in a statement, pointing to a separate ruling by Judge Rita Lin in the Northern District of California, which blocked the Trump administration's attempt to bar Anthropic from all federal contracts. "We remain confident in our position and are considering all options, including further review." [1]
The two rulings now sit in direct tension—one federal court blessing the ban, another blocking its broader cousin. That split doesn't resolve anything; it sharpens everything.
The procedural path forward is narrow. Because the Federal Acquisition Supply Chain Security Act routes cases directly to the DC Circuit, Charlie Bullock of the Institute for Law & AI explained, "the only appeal from there is to the Supreme Court (which doesn't have to accept appeals; it exercises discretion and generally grants only a small percentage of appeal requests) or to an en banc panel of the D.C. Circuit (also discretionary, also not granted super often)." [1]
Sean Timmons, a former JAG Corps officer now at law firm Tully Rinckey, noted that if a full panel is granted, "all bets are off." The full DC Circuit has a mix of Trump, Biden, Obama, and Bush appointees—far more ideologically diverse than the panel that ruled today. "You've got two people appointed by Trump [saying] we're not going to step in and undo a decision based on national security parameters, because national security is the sole prerogative of the chief executive," Timmons told Breaking Defense. "It's pretty egregious… but when you invoke national security, the courts are kind of like Pontius Pilate: They just want to wash their hands and walk away." [1]
Both Timmons and Bullock independently concluded that the case's profile could compel the full court to act. "This was the expected outcome, given the terrible luck Anthropic had with the panel draw," Bullock posted online. "I would expect Anthropic to win before an en banc DC Circuit and likely even before SCOTUS, but they're not guaranteed any appeal at all." [1]
What this case is really testing isn't Anthropic specifically—it's whether a government can classify a company's ethics commitments as a security threat. That argument has consequences well beyond any single AI model.
The immediate effect of the ruling is concrete: defense contractors using Claude must now find alternatives or risk contract violations. The longer legal arc, however, remains wide open. Anthropic's assembled legal team, flag-waving dissent from Judge Henderson, and a contradictory California ruling give the full DC Circuit—or the Supreme Court—ample reason to revisit the question. Whether either court chooses to do so is another matter entirely. [1]
When you invoke national security, the courts are kind of like Pontius Pilate: They just want to wash their hands and walk away.Sean Timmons