Judge blocks Trump administration's 'retaliatory' punishment of AI firm Anthropic

Late Thursday afternoon, a federal judge in San Francisco handed Anthropic a decisive victory in its monthslong clash with the Trump administration over artificial intelligence and national security. U.S. District Judge Rita Lin issued a permanent order blocking the government from enforcing a series of rules designed to cut the AI company off from federal contracts and military use. The judge found that the administration had violated the First Amendment by punishing Anthropic for protected speech—specifically, the company's public insistence on safety guardrails for its Claude AI model.

The conflict began over a fundamental disagreement about how Claude should be deployed. Anthropic, led by CEO Dario Amodei, wanted two absolute restrictions: the model could not be used for mass surveillance of Americans, and it could not power autonomous weapons capable of targeting people without human decision-making. The military pushed back, arguing that existing federal law and internal Pentagon policies already prohibited those uses, making Anthropic's additional safeguards unnecessary. When negotiations collapsed in February, the Trump administration responded with public fury. Defense Secretary Pete Hegseth called the company "sanctimonious." President Trump labeled it "radical left, woke." Within days, the administration issued orders to cut off Anthropic's access to federal systems and barred military contractors from using its technology for any purpose.

Judge Lin's 59-page opinion dismantled the government's legal rationale. The administration had designated Anthropic as a "supply chain risk"—a classification normally reserved for foreign adversaries who might sabotage critical U.S. systems. The government claimed it feared Anthropic might install a "kill switch" in its software to enforce its own policy preferences on military users. Lin called this concern "entirely unfounded," noting that Anthropic had provided evidence it lacks the technical ability to maintain backdoor access to its systems. She also pointed out the logical inconsistency: if the government truly believed Anthropic posed a sabotage threat, why did it continue seeking to collaborate with the company on advanced AI development?

What struck the judge most forcefully was the timing and manner of the administration's actions. Trump had announced the ban via Truth Social before any formal administrative process had begun. Hegseth's February social media post declaring that military contractors could conduct "no commercial activity with Anthropic" had triggered a broader boycott that extended even to non-military uses of the technology. Lin wrote that "the decision to publicly broadcast Anthropic's punishment via social media—even before the formal, administrative process of making the necessary findings to designate Anthropic a supply chain risk had begun—makes little sense except as an attempt to swiftly make a public example of Anthropic for daring to criticize the Administration." She concluded that the moves constituted "retaliatory act, taken without due process, meant to punish Anthropic for its protected speech."

The judge acknowledged that the government deserves deference on national security matters. But she drew a sharp line between legitimate security decisions and punishment disguised as policy. "The empty invocation of national security is not a blank check to punish and retaliate against government critics," she wrote. She also rejected the government's argument that it was simply exercising its right to choose which vendors to buy from. An IT vendor does not become an adversary of the United States merely by asking tough questions or insisting on particular contract terms, Lin reasoned, even if those demands frustrate federal officials.

Anthropicwelcomed the ruling, saying it remained committed to working with the government on AI for national security. The company had drawn support from civil liberties organizations and from Microsoft, which warned in a court filing that the supply chain designation could cause "severe economic effects that are not in the public interest." The Justice Department had argued that it cut off Anthropic not for its speech but because the Defense Department no longer trusted the company's judgment. Lin found that argument unconvincing given the contemporaneous statements from Trump and Hegseth attacking Anthropic's values and ideology.

The ruling does not require the government to use Anthropic or prevent it from losing business to competitors, so long as those decisions comply with law and the Constitution. But it does mean the administration cannot enforce the punitive measures it imposed. A separate challenge to the government's actions is still pending in the U.S. Court of Appeals for the D.C. Circuit. The broader question—how to balance the innovation benefits of AI against legitimate national security concerns, and who gets to decide—remains unresolved. For now, Anthropic has won the right to compete for federal business on equal legal footing, even as the deeper debate over AI governance and executive power continues.