A federal appeals court in Washington has refused to suspend the Pentagon’s supply-chain risk designation against Anthropic, leaving defense contractors with conflicting legal signals over whether they can continue using Claude, and putting the ruling at odds with a separate federal court that reached the opposite conclusion last month.

“The equitable balance here cuts in favor of the government,” a three-judge panel wrote in its order Wednesday. “On one side is a relatively contained risk of financial harm to a single private company. On the other side is judicial management of how, and through whom, the Department of War secures vital AI technology during an active military conflict.”

The panel, comprising Judges Henderson, Katsas, and Rao, acknowledged that Anthropic “will likely suffer some degree of irreparable harm” but found its interests “seem primarily financial in nature” rather than constitutional.

The order states the ruling is not a final decision on the merits. Oral arguments are set for May 19.

Anthropic had asked the US Court of Appeals for the District of Columbia Circuit to pause the supply-chain risk designation issued March 3 by Secretary of War Pete Hegseth.

The label, according to the company’s court filings, bars it from Pentagon contracts and requires defense contractors to stop using Claude in military work. The court denied the request, conflicting with a US District Court in California that granted Anthropic a preliminary injunction on March 26, blocking a parallel designation under a related statute.

Acting Attorney General Todd Blanche called the ruling “a resounding victory for military readiness” in a post on X. “Military authority and operational control belong to the Commander-in-Chief and Department of War, not a tech company,” he wrote.

Vendor risk is no longer predictable

For enterprises, the split ruling creates a compliance problem with no clean answer. The order states the Department has canceled its contracts with Anthropic, begun removing Claude from its systems, and prohibited contractors from using it as a subcontractor on Pentagon work. It also states, however, that “the Department has not prohibited contractors from using Claude for work performed for entities other than the Department.”

That distinction does not resolve the uncertainty. Following the California injunction, the government filed a compliance status report on April 6, cited in legal analysis by Herbert Smith Freehills Kramer, confirming it had restored Anthropic access across federal systems. That compliance applied only to the California statute. The broader D.C. designation remains active.

Sanchit Vir Gogia, chief analyst at Greyhound Research, said enterprises are dealing with vendor risk that their procurement frameworks were not designed to handle. “It means a vendor does not have a single legal status anymore. It can be restricted under one framework and protected under another, at the same time. That is a very different world from the one enterprise procurement teams are used to operating in,” he said.

The timing mismatch compounds the problem, Gogia said. “Legal processes move on their own timelines. Procurement cycles move on to another. Architecture decisions, once made, are not easy to reverse. When those timelines fall out of sync, you end up locked into dependencies that may no longer be viable,” he said.

‘Any lawful use’ shifts governance into the contract

The case has implications beyond Anthropic, Gogia said. The “any lawful use” standard the Pentagon sought to impose is one that the General Services Administration is separately moving to codify across federal AI procurement.

If that happens, governance authority would move from vendor-defined safeguards into contract language, Gogia said. “The contract becomes the final authority, not the platform. Governance is no longer primarily enforced through design. It is enforced through legal agreement,” he said.

Large defense contractors required to operate under such terms will push equivalent requirements down their supply chains, Gogia said, meaning enterprises with no direct Pentagon exposure may still face similar obligations through their partners.

On Anthropic’s refusal to drop its ethical restrictions, he said the question enterprises ultimately ask is “not whether a vendor is ethical, but whether that vendor can remain usable across all the contexts in which the enterprise operates.”

Matt Schruers, CEO of the Computer & Communications Industry Association, which filed an amicus brief in the case alongside ITI, SIIA, and TechNet, said the outcome adds to an already difficult environment. “The Pentagon’s actions and the DC Circuit’s ruling create substantial business uncertainty at a time when US companies are competing with global counterparts to lead in AI,” he said in a statement.

The D.C. court directed both parties to address three unresolved threshold questions before May 19, including whether the court has jurisdiction over Anthropic’s petition at all, according to the order. Anthropic’s opening brief is due April 22. Anthropic did not immediately respond to a request for comment.