A federal judge has struck down the Pentagon’s designation of Anthropic as a “supply chain risk” as unlawful. U.S. District Judge Rita Lin, presiding in California, deemed the label “arbitrary and capricious,” ruling that it violated both the First and Fifth Amendments by acting as unlawful retaliation and denying due process. The ruling came down Thursday evening, providing Anthropic its first judicial victory in the dispute.
The core of the disagreement began earlier in 2026, when Defense Secretary Pete Hegseth and President Donald Trump assigned Anthropic the supply chain risk label. Government agencies were thereafter ordered to sever contracts and partnerships with the AI company—even those outside the defense sector.
Why Anthropic was flagged
The government’s justification focused on Anthropic’s insistence on robust safety guardrails—specifically, guardrails that would prevent its models from being used for fully autonomous weapons systems or enabling mass surveillance of American citizens. The Pentagon argued that Anthropic could retain some control over how its AI models were used even after an agency purchased them.
Judge Lin rejected these claims. Her ruling emphasized that Anthropic demonstrably does not provide any “backdoor access” to its technology after transfer to the Department of Defense. She also noted contradictions in the government’s behavior—among them, proposals labeling the company essential to national security and ongoing collaborations involving its new model, Mythos, for cybersecurity efforts.
Constitutional concerns and government conduct
According to Judge Lin, the label amounted to retaliation against Anthropic for criticizing government policy. The decision also lacked the procedural fairness that the Fifth Amendment requires. The judge saw the administration’s invocation of national security as an empty pretext being used to target a critic rather than a genuine security concern.
Moreover, Lin questioned the government’s inconsistency—being willing to potentially designate Anthropic as essential under the Defense Production Act, while simultaneously treating it as a threat. She also pointed out that contracts and collaboration with Anthropic continued despite the label.
Anthropic has filed lawsuits in both California and Washington, D.C.; the California case was resolved with this ruling. The D.C. lawsuit remains active.
Anthropic responded by affirming its respect for the court’s decision, expressing continued intent to cooperate with federal agencies to ensure AI’s responsible use in domains tied to national security.
This marks a major narrow victory for Anthropic, but the underlying legal battles are far from over. With the D.C. suit underway and policy debates around AI regulation growing louder, the implications stretch well beyond this case. As courts begin weighing the limits of governmental power to label private AI firms, the line between national security and free political speech is becoming a hotly contested battlefield. Now, all eyes will be on how the government responds, whether through appeals or policy adjustments—and what that means for the balance between innovation, accountability, and constitutional rights.
