A three-judge panel of the U.S. Court of Appeals for the District of Columbia Circuit issued a 2-1 decision permitting the Pentagon to add Anthropic to a blacklist after the firm refused to turn on specific Claude functions. The ruling affirmed the department’s power to restrict a supplier whose technology it deems a security liability.
The two judges who voted against Anthropic were appointed during the previous administration, with one previously serving as deputy counsel to the president and the other having worked in the Office of Management and Budget. Their opinion warned that AI models that are overly limited could jeopardize missions, while unrestricted systems might generate false target data for lethal weapons.
The court distinguished the lower-court analysis, which relied on 10 U.S.C. § 3252, a statute that limits supply-chain risk designations to hostile actors. Instead, the appeals panel applied 41 U.S.C. § 4713, a broader provision that covers any person who might sabotage, deny, or otherwise compromise a technology, and which Congress has assigned exclusive review to the D.C. Circuit.
Anthropic had sued the administration in March after a directive ordered federal agencies to cease using its products and barred defense contractors from contracting with the company. The company’s emergency request for a stay was denied in April, and a California district judge later ruled that the blacklist violated the First Amendment, calling the national-security justification a “blank check.”
The ruling leaves Anthropic with two avenues for further challenge: it can request an en banc hearing before the full D.C. Circuit or petition the Supreme Court for review. Meanwhile, the court noted that President Trump and Defense Secretary Pete Hegseth must weigh the dangers of both over-constraining and under-controlling AI in military contexts.
Legal scholars see the decision as a signal that the government may invoke expansive supply-chain authority to manage emerging AI risks, even when the perceived threat does not stem from a foreign adversary. The judgment underscores the tension between protecting national security and preserving constitutional freedoms in the rapidly evolving field of artificial intelligence.