It began with a refusal. When Anthropic, the San Francisco company behind the Claude family of AI models, declined to lift its own restrictions on deploying its technology for lethal autonomous weapons and mass domestic surveillance, the Trump administration responded by designating the firm a supply-chain risk to national security — effectively cutting it off from federal contracts and defence work across the country. That decision has now been ruled unlawful. Judge Rita Lin of the US District Court for the Northern District of California vacated the government's directives in a ruling that granted the core of Anthropic's motion for summary judgment.

The administration had, as Lin noted, gone far beyond any reasonable vendor dispute: President Trump and Defence Secretary Pete Hegseth had "ordered all federal agencies to permanently stop using Anthropic's products, and banned all defence contractors from doing any business with Anthropic, even if unrelated to the military." Lin found that this broad campaign could not be squared with the First Amendment's protections against government retaliation for speech. The administration's justification, which had centred on the claim that Anthropic possessed a secret backdoor into its own technology once it was deployed in national security systems, collapsed under scrutiny. Defendants ultimately conceded that no such access existed and that Claude posed no greater risk than any other "black box" AI model.

Lin was direct in her assessment: "The empty invocation of national security is not a blank check to punish and retaliate against government critics." The ruling carries implications well beyond one company's contracts.