Judge rules Trump administration illegally punished Anthropic over AI safety refusals

The Trump administration violated Anthropic PBC's constitutional rights when Secretary of Defense Pete Hegseth labeled the company a national security supply chain risk over its refusal to allow the government to use its AI models for lethal autonomous warfare or mass surveillance, a judge ruled Thursday evening.

U.S. District Judge Rita F. Lin agreed that the federal government is entitled to deference on matters of national security but concluded that the actions by Hegseth and President Donald Trump were based on a desire to punish the San Francisco company for criticizing the government.

"Defendants' contemporaneous words and deeds confirm that the challenged actions were based on a desire to make a public example out of Anthropic for its 'arrogance' in criticizing the government, not based on any articulable basis to believe that Anthropic would actually sabotage its model," she wrote.

Lin granted Anthropic's motions arguing that the government violated its constitutional rights as well as its claims that Hegseth's directive violated the Administrative Procedure Act involving some federal agencies, denying them only for other agencies whose actions against the company were tentative.

The San Francisco judge denied the federal government's motions for summary judgment.

"The undisputed record shows that the challenged actions constituted unlawful retaliation in violation of the First Amendment, and that Anthropic was denied the pre-deprivation process required under the Fifth Amendment," Lin wrote.

"The empty invocation of national security is not a blank check to punish and retaliate against government critics," added Lin, an appointee of President Joe Biden. Anthropic PBC v. U.S. Department of War et al., 26-cv-01996 (N.D. Cal., filed March 9, 2026).

The judge blasted the government's assertions in its brief and during oral arguments last month that Anthropic's criticisms left it with legitimate concerns the company would sabotage its own models.

Senior Trial Counsel James W. Harlow of the Justice Department said during oral arguments that the government's concerns were legitimate and that "there should be deference to the judgment of these [Defense Department] officials."

Lin disagreed, siding with Michael J. Mongan, a partner with Wilmer Cutler Pickering Hale and Dorr LLP who represents Anthropic, which has worked with the government on a cybersecurity AI tool since its dispute with the Defense Department.

"Even now, the government is discussing collaboration with Anthropic on its new model, Mythos, in an array of sensitive contexts," she wrote. "None of that is consistent with a genuine fear that Anthropic is a saboteur who would poison its software to harm national security."

Anthropic is one of the nation's leading AI companies and is gearing up for an initial public offering, possibly as soon as this fall.

Lin denied the government's request that a ruling against it be stayed for seven days.

"Defendants have not shown irreparable harm, even though they have been complying with the Court's Preliminary Injunction Order for more than five months and had ample opportunity to identify any harms caused by it," she wrote.

The Justice Department could not be reached as of press time about whether the government would appeal Lin's decision to the 9th U.S. Circuit Court of Appeals.