Appeals court allows US blacklist of Anthropic over Claude

The U.S. Court of Appeals for the District of Columbia Circuit issued a 2-1 ruling affirming the government's authority to block Anthropic. The decision stems from Anthropic's refusal to lift safety restrictions on its Claude AI model, which prevented government users from employing the technology for lethal autonomous warfare and domestic mass surveillance. The majority opinion, authored by Trump-appointed Judges Gregory Katsas and Neomi Rao, concluded that the Department of Defense reasonably feared Anthropic might manipulate Claude to block authorized national security functions during critical operations.

This ruling hinges on 41 U.S.C. Section 4713, a broad procurement statute that defines supply chain risks as actions by "any person" that could disrupt or deny technology operations. The court decided this definition applies to Anthropic's safety guardrails, even without malicious intent. In contrast, Judge Karen Henderson dissented, arguing the law was meant to target hostile foreign actors infiltrating systems, rather than a contractor's, as she wrote, "honest and upfront enforcement" of its own terms of service.

The decision complicates an ongoing legal battle. Last month, Judge Rita Lin of the U.S. District Court for the Northern District of California ruled the blacklist illegal under a different statute, 10 U.S.C. Section 3252, which requires proof of an adversary's bad motive. Lin, a Biden appointee, also found that the ban violated the First Amendment, writing that national security concerns do not grant a "blank check" to retaliate against critics. However, the D.C. Circuit asserted exclusive jurisdiction over the broader Section 4713 statute.

Anthropic sued the administration in March after President Trump and Defense Secretary Pete Hegseth barred federal agencies and defense contractors from using its products. While Anthropic is considering further appeals, including a Supreme Court petition, Commerce Secretary Howard Lutnick recently indicated that the administration and the AI startup have improved relations and are now "in tune."