A US federal judge has ruled that the Pentagon illegally blacklisted Anthropic as a national security threat based partly on capabilities its AI system Claude didn't actually possess. US District Judge Rita Lin determined in a 59-page decision that the government unlawfully retaliated against the AI maker for speaking out, denied it due process, and wrongly labeled it a supply-chain risk. She called some of the Pentagon's core claims about Claude "entirely unfounded."
The conflict began in February when Anthropic refused to strip away protections that prevent Claude from being used for specific military purposes, including fully autonomous weapons and large-scale domestic surveillance. Defense Secretary Pete Hegseth gave the company three days to agree to "all lawful uses" or face designation as a supply-chain risk, despite calling Claude's capabilities "exquisite" and the company's worries "understandable." President Trump then ordered federal agencies to stop using Anthropic's technology. The Pentagon's assessment argued Anthropic could interfere with Claude during military operations by changing or shutting down models, letting them "drift," or inserting hidden biases or backdoors. But according to the court, Anthropic couldn't actually perform those actions on Claude models already running inside Pentagon systems, since those models are static and the company can't remotely access, modify, update, or disable them. The government didn't challenge that evidence. Years of government records showed no prior supply-chain concerns about Anthropic, the judge noted. More than 100 enterprise customers reached out to Anthropic worried about continuing their relationships, and the company estimated the measures could cut multiple billions of dollars from its 2026 revenue.
Lin wrote that the timeline indicated the Pentagon built its justification "after the fact to justify the foreordained conclusion." The judge found substantial evidence the company had been punished for publicly opposing the administration, citing Trump calling it a "RADICAL LEFT, WOKE COMPANY" staffed by "Leftwing nut jobs" and Hegseth's references to its "Silicon Valley ideology." The Pentagon's own internal memo complained Anthropic had acted in an "increasingly hostile manner through the press." Lin said revealing Anthropic's punishment before the formal designation process started appeared intended to "make a public example of Anthropic for daring to criticize the Administration." She rejected the notion that being a challenging supplier transforms a company into a security threat, writing that "an IT vendor does not become a potential adversary of the United States whenever it asks probing questions."
The ruling emphasized the Pentagon can lawfully stop purchasing Anthropic's technology through standard procurement channels, but it can't weaponize a contracting dispute to silence a government critic. Lin concluded that "the empty invocation of national security is not a blank check to punish and retaliate against government critics." The government's behavior also contradicted its stated threat assessment: even after finalizing the designation, Under Secretary of Defense Emil Michael continued negotiating contract terms with Anthropic, writing "I think we are very close here" in an email. The NSA and NASA began deploying Anthropic's Mythos cybersecurity model around Project Glasswing, while senior White House officials held what Anthropic described as "productive and constructive" discussions with CEO Dario Amodei. Hegseth's directive had sought to block Pentagon contractors and suppliers from working with Anthropic even on projects unrelated to defense, casting a shadow far beyond a single lost contract. The decision suggests government agencies must ground national security designations in verifiable facts rather than political disagreements, setting boundaries on how far officials can stretch security justifications when facing corporate pushback. Companies navigating high-stakes government relationships now face a clearer test: technical guardrails and public criticism may trigger political backlash, but courts will scrutinize whether security labels mask retaliation for dissent.