A U.S. federal judge rules that the Trump administration acted illegally when it designated AI company Anthropic as a “national security supply-chain risk,” a move that had barred federal agencies and contractors from using its Claude models. In a 59-page order, Judge Rita Lin issues a permanent injunction in the case filed in California, saying the government’s rationale is baseless and that it unlawfully targeted Anthropic for criticizing Defense Department views on AI safety and battlefield use.
The dispute stems from months of negotiations between Anthropic and the Pentagon over military access to Claude. Anthropic says it would not accept unrestricted use that would enable domestic mass surveillance or fully autonomous lethal weapons. After talks stalled, Defense Secretary Pete Hegseth imposed the supply-chain designation, invoking a procurement statute intended to protect national-security systems from sabotage or subversion. Multiple outlets report the court found the government’s central theory does not hold: the government later concedes Anthropic has no backdoor access to deployed models and that its models are not riskier than other “black box” systems.
Outlets also describe differing dimensions of the broader conflict. Several note First Amendment retaliation and due-process issues were central to Lin’s reasoning, while other reporting highlights related, separate government actions and additional litigation still pending in other courts. The government is expected to appeal, and parallel processes continue to shape how Anthropic’s models are treated for defense purposes.
