Federal judge rules Trump administration's Anthropic blacklist was illegal
A US federal judge ruled that the Trump administration illegally retaliated against AI company Anthropic by barring federal agencies and defense contractors from using its products, ordering the directives be rescinded.

A judge in the US District Court for the Northern District of California has ruled that the Trump administration acted unlawfully when it blocked federal agencies and defense contractors from using technology made by Anthropic, the company behind the Claude AI system. Judge Rita Lin granted key parts of Anthropic's motion for summary judgment, vacating the government's directives and ordering them to be rescinded.
According to the ruling, Anthropic had been designated a supply-chain risk to national security after the company refused to lift restrictions on using its products for lethal autonomous weapons and mass surveillance of American citizens. President Trump and Defense Secretary Pete Hegseth had ordered all federal agencies to permanently stop using Anthropic's products and barred any defense contractor from doing business with the company, even in matters unrelated to military work.
First Amendment violation cited
Judge Lin found that the evidence clearly showed the government's actions amounted to unlawful retaliation in violation of the First Amendment. She acknowledged that the Department of Defense remains free to choose whichever AI vendor it prefers, but concluded that the sweeping restrictions imposed on Anthropic were baseless and illegal.
The judge further stated that invoking national security does not give the government a blank check to punish critics. During proceedings, government representatives offered only a thin justification for their actions and ultimately abandoned their central claim — that Anthropic maintained backdoor access to its technology once deployed within national security systems. The court found no evidence of such access, and that Anthropic's technology poses no greater risk than any other "black box" artificial intelligence model.


