US federal judge permanently blocks Pentagon blacklisting of AI maker Anthropic

Court ruling and constitutional findings

US District Judge Rita Lin in the Northern District of California permanently blocked the Trump administration's directive designating Anthropic as a supply-chain risk to national security. In a 59-page decision, Lin granted key portions of Anthropic's motion for summary judgment, concluding that the government's blacklisting violated the First Amendment and Fifth Amendment due process protections. The judge found that the Department of Defense sought to punish Anthropic for refusing to remove safety guardrails from its Claude artificial intelligence model. The court rejected the government's claim that Anthropic posed a technical threat, noting that federal officials conceded the company lacked backdoor access to models deployed in national security systems.

The empty invocation of national security is not a blank check to punish and retaliate against government critics.

Origins of the autonomous weapons clash

The confrontation began in early 2026 after Anthropic refused Pentagon requests to eliminate restrictions on how Claude could be deployed. Anthropic had included provisions in its contracts prohibiting the use of its technology for lethal fully autonomous weapons and mass domestic surveillance of American citizens. The company argued that generative models remained insufficiently reliable to make life-or-death battlefield decisions without human intervention. Friction intensified following a January military operation to capture Venezuelan President Nicolás Maduro, after an Anthropic executive inquired whether Palantir had deployed Claude during the mission. Pentagon officials demanded full operational authority for all lawful purposes, arguing that private vendors cannot veto decisions made by elected officials.

Escalation across federal agencies

On 23 February 2026, Defense Secretary Pete Hegseth summoned Anthropic chief executive Dario Amodei to the Pentagon to demand unrestricted access. Pentagon spokesperson Sean Parnell subsequently set a deadline of 5:01 p.m. ET on 27 February for the company to comply. Following Anthropic's refusal, President Donald Trump directed every federal agency to immediately cease using Anthropic software, and Hegseth initiated supply-chain risk designations. The Departments of State, Treasury, and Health and Human Services moved to halt use of Claude in early March, while defense contractors including Lockheed Martin and Boeing were instructed to assess or end their ties. Meanwhile, the Defense Department continued working with Anthropic on cybersecurity tools through its Mythos model and considered applying the Defense Production Act, which Judge Lin cited as evidence contradicting the government's threat designation.

Commercial stakes and ongoing litigation

The ruling vacates the government directives and orders the administration to rescind all instructions blacklisting Anthropic. The decision arrives as Anthropic prepares for a planned initial public offering targeting a valuation of approximately $2 trillion. An Anthropic spokesperson stated that the company remains focused on working with the government to harness artificial intelligence for national security. However, the legal dispute is not entirely resolved. The Trump administration retains the right to appeal Judge Lin's ruling to a higher court, while a second related lawsuit under federal procurement statutes remains pending before a federal court in Washington.

We remain focused on working productively with the government to harness AI for our national security so all Americans benefit from this technology.