A US judge has said the Trump administration has not yet shown sufficient evidence to classify AI company Anthropic as a supply-chain risk and bar federal agencies from using its technology. The case matters because it tests whether the government can penalise contractors over disagreements about military uses of AI, potentially affecting free speech and procurement rules.

The dispute followed failed Defence Department contract talks after Anthropic opposed uses involving mass surveillance of Americans and lethal targeting or firing decisions, saying its technology was not ready. Judge Rita Lin also questioned claims that Anthropic could change or disable delivered models during military operations, finding no evidence of a “kill switch”; she had temporarily blocked the ban in March and is deciding whether to make that order permanent.