Court rules Pentagon can’t ban Anthropic’s AI models

A federal court has knocked out the Defense Department’s ban on Anthropic’s AI models, calling the move “illegal and baseless” in a late Thursday ruling from Judge Rita Lin. The case started when the Pentagon tried to get Claude onto an internal system called GenAI.mil, then objected to a clause in Anthropic’s terms that bars mass surveillance of Americans and autonomous weapons use. The department wanted that language replaced with a blanket allowance for “all lawful uses” of Claude. Anthropic refused. That refusal set off a wider fight, with President Donald Trump banning federal agencies from using the models and Defense Secretary Pete Hegseth branding Anthropic a supply chain risk, which also blocked defense contractors from using Claude. Anthropic sued in March, and the court sided with the company after rejecting the Pentagon’s two main arguments. One was that Anthropic had some kind of backdoor access to deployed national security systems. Officials later conceded it did not. The judge said Anthropic’s technology was no riskier to national security than any other black box AI model. The second argument was more political: that Anthropic had been openly hostile to the administration’s AI views. Judge Lin wasn’t having that either, writing that the Constitution and the federal statute cited by the government do not permit sweeping penalties just because a company criticizes the administration. The order also pointed out that the Pentagon’s own AI usage policies did not line up neatly with its lawsuit stance. There’s still a procedural catch. Anthropic also filed a complaint with the U.S. Court of Appeals for the D.C. Circuit, and according to CNBC, that judge will also need to rule for the company before the Claude ban is fully gone. Anthropic, meanwhile, says the designation was unlawful and says it still wants to work with the government on national security uses of AI.