Judge voids Pentagon blacklist of Anthropic as illegal retaliation

U.S. District Judge Rita Lin ruled on Thursday that the Pentagon's decision to brand Anthropic a supply-chain risk was unlawful. The 59-page opinion, filed in the Northern District of California, vacates Defense Secretary Pete Hegseth's 27 February designation and lifts a second order that barred military contractors from dealing with the company behind Claude.

Lin wrote that the empty invocation of national security is not a blank check to punish and retaliate against government critics. She found that the broad measures imposed on Anthropic were illegal and baseless. The Department of War, she added, remains free to pick any AI vendor it wants. What it may not do is use an obscure procurement statute to cripple a U.S. firm for public criticism of how the military planned to use its models.

The fight began in February. President Donald Trump and Hegseth accused Anthropic of endangering national security after Chief Executive Dario Amodei refused to loosen limits on uses that the company said could enable mass surveillance or autonomous armed drones. The Pentagon then listed Anthropic as a national-security and supply-chain risk, a label that can freeze federal work and scare off commercial partners. Anthropic sued.

Lin, appointed by former President Joe Biden, had already paused the penalties in March. Thursday's order makes that pause permanent for the named defendants. She declared that the challenged actions violate the First Amendment as retaliation for protected speech and violate due process. Nine agencies, among them the Pentagon, the Treasury, the State Department and the Department of Homeland Security, had piled on extra sanctions. Those penalties fall with the order.

The statute the government used was written to keep foreign sabotage out of military systems. Anthropic argued that the label had never before been applied in public to an American company. The firm said the designation risked billions in lost contracts and lasting reputational harm. Lin accepted the core of that account. She wrote that the government's actions were based on a desire to make a public example of Anthropic for what officials called arrogance in criticizing the administration, not on any articulable basis to believe the company would sabotage its model.

The ruling does not force the Pentagon to buy Claude. It does not restore every commercial relationship that cooled after February. It does restore Anthropic's ability to sell to civilian agencies and to work with vendors who had been told to stay away. A related challenge under a different statute is still moving in the D.C. Circuit. The administration is expected to appeal the California judgment.

CourtListener posted the accompanying order of final relief in case 26-cv-01996-RFL. Anthropic's motion for summary judgment was granted in part and denied in part. Defendants received the same mixed result. The First Amendment claim succeeded. Permanent injunctive language reaches officers, employees, successors and anyone acting in concert with them.

The case sits inside a wider argument about who sets the rules for military AI. Amodei has said the company will sell to the government under conditions. The Pentagon has said it will not let a vendor dictate terms for battlefield systems. Lin's opinion does not settle that policy fight. It says the government cannot convert a contract preference into a public punishment dressed as a supply-chain finding.

For other AI vendors watching the docket, the practical takeaway is narrower than the headlines. A cabinet department can still walk away from a model. It can still write its own acceptable-use rules. What it cannot do, on this record, is invoke an anti-sabotage tool against a domestic firm whose main offense, in the court's view, was saying no in public.

Friday's coverage in Axios, Wired, the Los Angeles Times and the Guardian tracked the same quotes from page after page of the opinion. None of those outlets produced a government comment on the record. A Pentagon spokesperson could not be reached immediately after the filing. The legal clock now runs on appeal, while Anthropic returns to selling Claude where the injunction allows.

The next test is whether the D.C. case and the Ninth Circuit appeal produce a split that the Supreme Court would have to resolve. Until then, the supply-chain label on Anthropic is void in the district that issued it, and the nine named agencies are barred from giving it effect.