A California federal judge has ruled that the US Department of Defense unlawfully retaliated against the artificial intelligence company Anthropic, in a decision that could shape how far governments can push frontier AI firms. The court found the Pentagon had wrongly designated the company a supply-chain risk.
The dispute began after Anthropic declined a defence contract that lacked safeguards against lethal autonomous weapons and domestic mass surveillance. Chief executive Dario Amodei had maintained firm limits on how the company’s Claude models could be used for military purposes.
The ruling matters well beyond the two parties. It establishes a legal precedent that leading AI developers can resist government demands that conflict with their own safety policies, rather than being forced to choose between their principles and access to lucrative public contracts.
For an industry still writing its own rules, the decision offers a measure of protection for companies that set boundaries on how their technology is deployed. It suggests that safety commitments made publicly can carry legal weight when the state pushes back.
The matter is not fully settled. Additional litigation remains pending in Washington, and the government may appeal, meaning the precedent could yet be tested or narrowed by higher courts.
Still, the outcome is a notable moment in the relationship between the AI industry and the state. As governments race to harness the technology for defence and security, the ruling signals that the companies building it will not always be willing partners, and that the courts may back them when they refuse.
