A federal judge in California has blocked the U.S. Department of Defense’s attempt to designate Anthropic, a leading artificial intelligence lab, as a supply-chain risk over the company’s restrictions on military applications of its technology. The ruling, issued late Thursday by Judge Rita Lin, marks a significant legal victory for Anthropic amid ongoing disputes with the Pentagon.
The conflict originated earlier this year when Anthropic declined to enter into an open-ended contract with the Department of Defense due to its commitment to two ethical “red lines”: prohibiting the use of its AI systems for lethal autonomous weapons and domestic mass surveillance. These limitations reportedly drew sharp criticism from then-Defense Secretary Pete Hegseth and former President Donald Trump.
In response, the Trump administration designated Anthropic a supply-chain risk in February, a label typically reserved for entities linked to hostile foreign powers. This move prompted a federal ban requiring all government agencies to cease use of Anthropic’s products and effectively barred defense contractors from engaging with the company. Despite the ban, a transition period allowed the Pentagon to continue utilizing Anthropic’s technology temporarily.
Judge Lin’s decision characterized the administration’s action as “unlawful retaliation” and emphasized that invoking national security does not grant unchecked authority to penalize dissenting government contractors. The ruling permits federal agencies outside the Department of Defense to maintain access to Anthropic’s AI offerings. Several government branches have reportedly been testing Anthropic’s Mythos model, leveraging its capabilities to assess cyber and national security risks.
The court’s judgment follows a separate legal matter involving Anthropic currently before a Washington appeals court. In April, that court rejected the company’s request to suspend the supply-chain risk designation. The government may choose to appeal the California ruling as well.
Despite tensions earlier in the year, relationships between Anthropic and parts of the administration have shown signs of improvement. The company collaborated with officials on the cautious rollout of its Mythos model, which has been delayed due to concerns about potential misuse of its cybersecurity features. In June, the government briefly blocked another Anthropic model, Fable, before reaching an agreement allowing its eventual release.
Anthropic expressed satisfaction with the court’s decision, affirming its intent to work constructively with the government to develop AI technologies that support national security while benefiting the broader public. Neither the White House nor the Department of Defense provided comment on the ruling.