A 59-page order issued Thursday night by U.S. District Judge Rita Lin has undone one of the most aggressive moves Washington has made against an American AI developer. Lin found that the Pentagon's decision to brand Anthropic a supply chain risk was unlawful retaliation under the First Amendment. She also found the company was denied the pre-deprivation process the Fifth Amendment requires, and she ordered the designation removed.
Lin sits in the Northern District of California and was appointed by former President Joe Biden. Her order does not tell the Defense Department which vendors it must buy from. She was explicit that the military retains wide latitude over procurement. What she rejected was the machinery used to punish a critic.
How a procurement label became a speech case
The dispute traces back to February. Anthropic declined to strip internal safety guardrails that block its Claude models from being used in fully autonomous weapons and mass surveillance programs. The company argued its systems were not reliable enough for those roles.
Defense Secretary Pete Hegseth took the position that a private vendor could not dictate how the American military uses its tools. The administration then designated Anthropic a supply chain risk. It was an unprecedented application of the label.
The practical effect was near-total exclusion. No part of the Pentagon, including its contractors, could work with Anthropic products. Until then the designation had been reserved for firms viewed as tied to foreign adversaries. Anthropic sued in March, describing the move as an unlawful campaign of retaliation.
What the evidence showed
Lin's language was unusually direct for a procurement dispute. She wrote that the empty invocation of national security is not a blank check to punish and retaliate against government critics.
The record, she found, showed the Pentagon wanted to make a public example of Anthropic for what officials characterized as arrogance in criticizing the government. She saw no articulable basis for believing the company would actually sabotage its own model.
She also pointed to an inconsistency in the government's conduct. Other parts of the federal government kept meeting and working with Anthropic after the Pentagon acted. That pattern, in her reading, is not what agencies do when they genuinely believe a supplier might poison its software.
The signal had been building for weeks. At a July 30 hearing, Lin called the government's position troubling and said it appeared to sit at odds with the First Amendment. She noted the record had gotten worse for the government over time. Justice Department lawyers had argued that AI models are so enormous and opaque that the department cannot evaluate them the way it evaluates hardware. Anthropic attorney Michael Mongan countered that the measures threatened to chill debate on a question of real public importance.
The fight is not over
An Anthropic spokesperson said the company welcomes the ruling and remains focused on working productively with the government on national security applications. The government is expected to appeal. The White House and the Pentagon did not immediately respond to requests for comment.
A second, narrower case remains pending before the federal appeals court in Washington, D.C. It concerns a different rule the Pentagon has invoked to reach the same designation. Until that matter resolves, the label has not fully disappeared from Anthropic's record.
The competitive backdrop is hard to ignore. OpenAI struck its own Pentagon agreement within hours of the action against Anthropic. Both companies are now positioning for closely watched public offerings, which makes any federal risk designation a live item for prospective investors rather than an abstract policy fight.
For the broader industry, the ruling establishes something that did not exist before Thursday. A frontier lab that refuses a government use case on stated safety grounds now has a documented precedent that punishing the refusal carries constitutional limits. Whether that precedent survives appeal is the question every general counsel in the sector will be watching.
