Appeals Court Skeptical as Anthropic Fights Pentagon's Supply-Risk Label

A federal appeals court in Washington showed clear skepticism of Anthropic's bid to reverse the Pentagon's designation of the company as a supply-chain risk to U.S. national security. At oral arguments on May 19 before the U.S. Court of Appeals for the District of Columbia Circuit, two of the three judges on the panel pressed Anthropic's attorney on whether Defense Secretary Pete Hegseth had actually violated any law when he issued the designation in March, a declaration that led directly to a ban on government procurement of Claude AI technology.

How the Dispute Began

The original designation came in March 2026, at a point when the Pentagon and Anthropic had been in discussions over how Claude would be deployed by military personnel and contractors. Hegseth issued the supply-chain risk label without, in Anthropic's view, following the proper administrative procedures required for such a declaration. The company filed suit arguing that the secretary overstepped his authority and denied Anthropic the notice and opportunity to respond that administrative law typically requires. For background on the original Pentagon supply-chain dispute and what triggered it, see our earlier reporting from April.

The ban that followed the designation has concrete commercial consequences. Federal agencies cannot procure Claude services while the label stands, and government contractors working on sensitive programs are effectively blocked from one of the leading AI platforms. Several AI products from other vendors, including offerings from OpenAI and Google, remain available to federal buyers. For a company pursuing a valuation approaching $900 billion, the government AI market represents a substantial slice of potential revenue that is currently off limits.

The Judges' Questions

At the hearing, the panel's questions focused less on whether Anthropic actually poses a supply-chain risk and more on the threshold question of whether courts can review a Cabinet secretary's supply-chain security designation at all. Under administrative law, certain executive branch decisions, particularly those touching on national security, receive substantial deference and are sometimes insulated from judicial review entirely. The judges appeared particularly skeptical of Anthropic's argument that the designation was made "illegally," pressing the company's attorneys on the specific statutory hook that would give the court authority to second-guess the secretary's judgment.

That is a narrow but consequential framing. If the appeals court rules that the designation is not subject to judicial review, Anthropic's case fails without the court ever evaluating whether Hegseth was factually correct about supply-chain risk. The company's attorneys maintained that even decisions carrying a national security label must follow a minimum procedural floor, and that Anthropic received neither advance notice nor a meaningful chance to respond before the designation became effective. Whether a three-judge panel already leaning skeptical accepts that argument may be clear within weeks.

"Two of the three judges on the panel peppered Anthropic's attorney with questions about his claim that Defense Secretary Pete Hegseth made the declaration illegally in March, following a dispute over how the company's Claude AI chatbot would be used by the military." Bloomberg, May 19, 2026

Broader Stakes for AI and Government

The outcome will shape how AI companies navigate federal procurement at a moment when the Defense Department is among the largest potential buyers of AI services in the world. Anthropic has publicly stated its ambition to work with the federal government on AI safety and deployment standards, which creates an awkward dynamic: the company that has argued most publicly for responsible AI governance is currently barred from selling to the government that sets AI policy. Anthropic's commercial momentum in the enterprise market has so far offset the loss of federal business, but the situation is not indefinitely sustainable if the company wants to build durable government partnerships.

The case also sets precedent for how AI companies can respond when executive branch agencies act adversely without warning. Regulatory frameworks in Europe build in formal appeal mechanisms for AI companies facing adverse determinations, but U.S. administrative law gives courts a narrower basis for reviewing Cabinet-level security decisions. Whatever the DC Circuit rules, the AI industry will read the decision carefully. A ruling that executive supply-chain designations are effectively unreviewable would leave other labs with little recourse if they receive similar treatment.