A clash over who gets to set the limits on military AI has ended, for now, with a sharp judicial rebuke of the Pentagon. A federal court found that national-security language was used to punish Anthropic for refusing to abandon its safety lines.

The dispute began this winter, when the Defense Department sought authority to use AI systems for any lawful purpose. Anthropic accepted military and national-security work but refused to permit two uses: mass surveillance of Americans and fully autonomous weapons. Other major AI companies agreed to the Pentagon’s revised terms, while Anthropic held its ground.¹

The standoff escalated in February, when President Donald Trump and Defense Secretary Pete Hegseth accused Anthropic of threatening national security and designated it a supply-chain risk. The label could have cut the Claude maker off from federal business and military contractors. Anthropic sued in March, arguing the move was retaliation for its public criticism and its insistence on guardrails.²

On Thursday, U.S. District Judge Rita Lin struck down the designation in a 59-page ruling. She concluded that the Pentagon’s broad penalties were not grounded in evidence that Anthropic would sabotage its products. “The empty invocation of national security is not a blank check to punish and retaliate against government critics,” Lin wrote.³

Lin also highlighted the contradiction at the center of the government’s case: officials continued seeking Anthropic’s services, including cybersecurity collaboration, even while depicting the company as a security threat. That conduct, she found, was inconsistent with a genuine fear of sabotage.

Anthropic welcomed the decision and said it remained committed to productive work with Washington. The Pentagon had argued that opaque, complex AI models require unusual scrutiny and maintained it sought only lawful uses. The government is expected to appeal, while a separate Anthropic challenge to another Pentagon designation remains pending in Washington, D.C.¹