US appeals court upholds Pentagon blacklist of AI lab Anthropic

A US appeals court has upheld the Pentagon's decision to place AI developer Anthropic on a national security blacklist, affirming its designation as a supply-chain risk and closing a months-long standoff over military access to its frontier models.

The conflict traces back to late January, when the Defense Department pressed Anthropic to strip safety guardrails from its Claude models — safeguards the company said could otherwise allow autonomous weapons targeting and domestic surveillance.

The pressure escalated through February. Defense Secretary Pete Hegseth summoned Anthropic chief executive Dario Amodei to the Pentagon for talks on military use of Claude, after which the department asked the company to cooperate or face consequences, including a supply-chain risk label. Contractors including Boeing and Lockheed Martin were asked to assess their reliance on Anthropic's technology, and a Pentagon spokesperson set a deadline of 5:01 pm ET on February 27 for unrestricted military use.

Anthropic refused to remove the safeguards. President Donald Trump then directed every federal agency to halt use of the company's technology, Hegseth instructed the Defense Department to designate Anthropic a supply-chain risk, and Anthropic said it would challenge the move in court. OpenAI separately announced a deal to deploy its technology in the department's classified network under three conditions.

In early March, the State, Treasury and Health and Human Services departments moved to stop using Claude, and Lockheed Martin said it would follow the Pentagon's direction — a signal, legal experts said, of a wider contractor exodus. A big tech industry group urged de-escalation, warning that the designation created uncertainty and could limit the military's access to the best products. The Defense Department formalised the supply-chain risk designation on March 5, and the General Services Administration terminated Anthropic's OneGov deal a day later.

Anthropic sued on March 9 to block the blacklisting, calling it unlawful and a violation of its free speech and due process rights. Executives said the designation could cut 2026 revenue by billions of dollars and damage its reputation. Microsoft backed the lawsuit, warning of costly supplier disruptions.

An internal memo showed the Pentagon telling senior leaders that Anthropic's tools could remain in use beyond a phase-out if deemed critical to national security. Anthropic sought a stay from the US Court of Appeals for the District of Columbia Circuit, while the Pentagon's chief technology officer, Emil Michael, ruled out negotiations. The administration argued the company was unlikely to succeed on its First Amendment claims.

At a March 24 hearing, US District Judge Rita Lin of the Northern District of California said the blacklisting "looks like an attempt to cripple Anthropic" over its AI safety concerns. Two days later she temporarily blocked the designation, giving the administration seven days to appeal. The appeals court has now upheld the Pentagon's action.