A federal judge on Aug. 27 permanently blocked the Defense Department’s (DOD) “supply-chain risk” designation of Anthropic, delivering a legal victory to the artificial intelligence (AI) company amid an ongoing dispute over military access to its frontier AI models.
U.S. District Judge Rita F. Lin granted Anthropic summary judgment on its First Amendment, Fifth Amendment due process, and Administrative Procedure Act (APA) claims against the federal government, rejecting the designation that had followed a months-long dispute between the company and the Trump administration.
DOD sought to use Anthropic’s Claude AI model for “all lawful purposes,” including sensitive military and intelligence applications. Anthropic objected to two uses: mass surveillance and fully autonomous weapons. As a result, President Donald Trump directed federal agencies to sever ties with Anthropic and Defense Secretary Pete Hegseth designated the company a supply-chain risk. Anthropic subsequently filed two lawsuits alleging unlawful retaliation, while the administration defended the designation as a lawful national security measure.
“Though the Department of War is undisputedly free to select the AI vendor of its choice, the evidence demonstrates that the broad measures imposed on Anthropic were illegal and baseless,” Lin wrote. “The empty invocation of national security is not a blank check to punish and retaliate against government critics.”
Under the Trump administration, the DOD was rebranded as the Department of War.
The federal government argued that Anthropic posed a supply-chain risk in part because the company could retain backdoor access to its technology after it was deployed in a national security system.
Lin rejected that rationale, finding that the Pentagon’s stated national security concerns were inconsistent with the administrative record. Her order said Anthropic did not have backdoor access to models deployed in national security systems.
“Anthropic’s technology is itself no riskier to the national security than any other ‘black box’ artificial intelligence model,” Lin wrote.
The government also pointed to what it called Anthropic’s “increasingly hostile manner through the press” and criticism of DOD’s views on AI use, arguing that officials “cannot trust Anthropic to ensure the integrity of its models.”
Lin rejected that justification as well, finding that the government could not penalize Anthropic for criticizing the administration.
“Neither the Constitution nor the federal statute invoked by Defendants allows them to impose sweeping penalties based principally on Anthropic’s critique of the Administration’s views,” Lin wrote.
She went further, finding that the record showed the government sought to punish Anthropic for its criticism rather than address a demonstrated security threat.
“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,” Lin wrote.
Lin also pointed to what she characterized as contradictions between the government’s stated security concerns and its dealings with Anthropic.
Days before designating Anthropic a supply-chain risk, Hegseth proposed invoking the Defense Production Act to compel Anthropic to provide its technology – a step Lin said would have treated the company as essential to national security rather than a threat to it.
Lin also noted that the government has continued discussing potential collaboration with Anthropic on its new Mythos model for sensitive applications.
“None of that is consistent with a genuine fear that Anthropic is a saboteur who would poison its software to harm national security,” Lin wrote.
She concluded that the government’s actions amounted to unlawful retaliation under the First Amendment and that Anthropic did not receive the notice and opportunity to respond required by the Fifth Amendment.
The ruling does not require the Pentagon to use Anthropic’s products or prevent it from transitioning to another AI provider. The court said the department remains free to end its relationship with Anthropic through actions that comply with applicable laws, regulations, and constitutional requirements.
The ruling extended beyond the Pentagon, with Lin granting Anthropic summary judgment on an APA claim challenging final orders from eight other agencies to terminate their use of Anthropic products.
Those agencies include the Treasury Department, Federal Housing Finance Agency, State Department, General Services Administration, Office of Personnel Management, Nuclear Regulatory Commission, Department of Homeland Security, and Energy Department.
The government prevailed on the same claim involving agencies where the record did not establish a final agency action. Those included the departments of Health and Human Services, Commerce, and Veterans Affairs, as well as the Securities and Exchange Commission and NASA.
Lin also denied the government’s request for a seven-day administrative stay. She found that the government had not demonstrated irreparable harm after operating under a preliminary injunction for more than five months.
A parallel case remains pending
The California ruling does not end Anthropic’s legal fight with the federal government, with a separate case over the Pentagon’s restrictions still pending in Washington, D.C.
In that case, a three-judge panel of the U.S. Court of Appeals for the District of Columbia Circuit declined to halt the Pentagon’s restrictions while the litigation proceeds. That outcome contrasts with the preliminary injunction Lin issued in California before permanently blocking the supply-chain risk designation.
The differing outcomes could ultimately set up further appellate review, potentially including by the U.S. Supreme Court.
DOD moves ahead with transition away from Anthropic
While Lin’s ruling blocks the supply-chain risk designation, it does not require the Pentagon to continue using Anthropic technology – and DOD officials said before the ruling that the department was already well into its transition.
DOD Chief Technology Officer Emil Michael said in June 2026 the department has completed roughly two-thirds of its transition from Anthropic-based infrastructure to competing AI providers.
Under the president’s directive, defense components and defense industrial base partners must remove all Anthropic software from their networks by Sept. 29.
Michael said alternative platforms, including OpenAI and Google Gemini, are being deployed across classified networks with minimal disruption.