Court: Pentagon Can't Punish AI Maker for Refusing Weapons Work

Anthropic, the company behind the Claude AI assistant, just won an important legal victory. A federal judge in California ruled that the Trump administration's decision to label Anthropic a national security risk was unlawful. That label had blocked all federal agencies from working with the company. The judge called it "unlawful retaliation" against Anthropic for criticizing the government.

What started this fight? Anthropic set firm limits on how the Pentagon could use its technology. The company refused to allow its AI to control fully autonomous weapons or be used for mass surveillance of American citizens. In response, Defense Secretary Pete Hegseth and President Trump placed Anthropic on a supply-chain risk list, which is normally reserved for foreign threats or security dangers. The judge found the evidence showed the opposite — the Pentagon was actually still pursuing contracts with Anthropic and even collaborating on cybersecurity.

The judge also shot down the government's claim that Anthropic posed a security risk. She noted that Anthropic cannot secretly access or control its technology after selling it to the Defense Department. The real motive, she wrote, was to make an example of Anthropic for its "arrogance" in criticizing the government's plans. As she put it, "The empty invocation of national security is not a blank check to punish and retaliate against government critics."

This ruling matters beyond one company. It tells other tech firms that they can refuse dangerous or unethical AI uses without fear of government punishment. It also reinforces that Americans' tax dollars shouldn't be used to silence companies simply for taking a moral stand. A second lawsuit from Anthropic in Washington, D.C., is still ongoing.