Judge Rules Pentagon's Anthropic Blacklist Was Illegal Retaliation

A federal judge has thrown out the Pentagon's attempt to brand Anthropic a "supply-chain risk," calling the designation illegal, baseless, and retaliatory. US District Judge Rita Lin issued the 59-page ruling on August 28, 2026, making permanent a suspension of the sanctions she had first ordered in March.

The dispute traces back to a $200 million contract for military use of Anthropic's Claude models. Anthropic refused to let its AI be deployed for autonomous weapons or domestic mass surveillance, arguing the models weren't reliable enough for lethal use and that the surveillance terms would violate constitutional rights.

Defense Secretary Pete Hegseth responded on February 27, 2026, by designating Anthropic a supply-chain risk and directing federal agencies to cut ties with the company. It marked the first time a US firm had been hit with that label, a status previously reserved for foreign companies under procurement law. President Trump amplified the move on social media, calling Anthropic "a radical left, woke company" and "out-of-control."

Judge Lin found the designation was "arbitrary and capricious" and violated Fifth Amendment due process protections. She also ruled it amounted to unlawful retaliation against Anthropic's public criticism of Pentagon AI policy, a First Amendment violation.

Judge Lin found the government had retaliatory intent tied directly to Anthropic's public stance on AI safety.

The government had argued the designation was purely contractual, stemming from Anthropic's refusal to accept terms permitting unrestricted military use of its models, not punishment for its safety positions. Lin rejected that account, tying the sanctions to Anthropic's public criticism rather than any documented supply-chain vulnerability.

The order vacates Hegseth's decision and bars nine federal agencies, including the Pentagon, Treasury, State Department, and Department of Homeland Security, from enforcing it. The ruling took effect immediately.

Anthropic executives had warned the label could cost the company billions in lost federal business and lasting reputational damage, given how procurement officers and other agencies treat a supply-chain risk flag as a red light. Anthropic had sued in March in both California and Washington, D.C., citing the government's actions as a coordinated retaliation campaign following its refusal to loosen restrictions around a Pentagon contract involving a model internally called Mythos, intended for cybersecurity work.

Lin's ruling doesn't force the Pentagon to work with Anthropic. She explicitly affirmed the department's discretion to choose whichever AI vendor it prefers, meaning the win is about removing an unlawful blacklist rather than compelling new business.

A second case remains pending in Washington, D.C., over a related supply-chain designation that could still block Anthropic from civilian government contracts. That case will test whether Thursday's reasoning extends beyond the Defense Department to the rest of the federal government.