The Pentagon’s clash with Anthropic was billed as a national-security dispute. A federal judge concluded it was something more pointed: an attempt to punish an AI company for drawing lines around how its technology could be used.

The confrontation began over the winter, when the Defense Department sought terms allowing use of AI models for “any lawful use.” Anthropic accepted military and intelligence work in principle but refused to permit mass surveillance of Americans or fully autonomous lethal weapons. Other major AI companies accepted the broader arrangement, while Anthropic held its ground.¹

The Trump administration responded by branding Anthropic a supply-chain risk, ordering federal agencies to stop using its products and barring defense contractors from doing business with it. The White House cast the company as a threat to military authority, saying President Trump would not let a “radical left, woke company” dictate how the armed forces operate.²

Anthropic sued in March, calling the designation retaliatory. Judge Rita Lin initially froze the blacklist; on Thursday, she delivered the company a fuller victory, finding the designation unconstitutional, arbitrary and capricious, and imposed without due process. Her 59-page decision found the government’s own conduct undercut its stated rationale: the Pentagon continued pursuing work with Anthropic and the company lacked the supposed backdoor access to deployed systems.³

“The empty invocation of national security is not a blank check to punish and retaliate against government critics,” Lin wrote.³ Anthropic said it welcomed the ruling and remained committed to working with government on national security.⁴

The administration is expected to appeal, and a separate Anthropic challenge to another Pentagon designation remains pending in Washington. The immediate blacklist is gone; the broader argument over whether AI vendors can impose safety limits on military customers is not.