Court Backs Pentagon Blacklist of Anthropic’s Claude
3 min readA federal appeals court has ruled that the Pentagon may keep Anthropic’s Claude out of military systems, and the reason is the model’s own safety limits. The decision turns a deliberate design choice into a procurement disqualifier.
How Claude Ended Up on a Blacklist
On March 3, Defense Secretary Pete Hegseth signed a determination labeling Claude a significant national security risk. The trigger was not a breach or a technical failure. Anthropic had declined to lift the restrictions in its usage policy covering lethal autonomous weapons and mass domestic surveillance, even after loosening several other safeguards to accommodate defense work.
The Pentagon acted under the Federal Acquisition Supply Chain Security Act of 2018, a law written to keep compromised hardware and software out of government networks. Anthropic sued, arguing the designation was retaliation for a policy position and that it violated the company’s First Amendment and due process rights.
What the Court Decided
On September 25 the U.S. Court of Appeals for the D.C. Circuit upheld the designation in a 2 to 1 ruling. Judges Gregory Katsas and Neomi Rao formed the majority. Judge Karen LeCraft Henderson dissented.
The majority held that the 2018 statute reaches operational risk regardless of a supplier’s intent. Because Anthropic designs Claude to refuse certain requests, the court reasoned, the Pentagon had enough grounds to treat the model as unreliable for military use. The panel rejected the constitutional claims.
Anthropic said it respectfully disagrees and is considering all options, including further review. The company pointed to an August federal district court ruling in California that struck down a separate, parallel designation under different procurement law. That decision is reportedly still in force, leaving the two cases pointing in opposite directions.
Why the Anthropic Pentagon Blacklist Matters
This is the first appellate decision to treat an AI model’s refusal behavior as a supply chain risk in its own right. If it stands, safety guardrails carry a measurable commercial cost, and every lab selling into government has to price that in. Rivals that ship fewer hardcoded refusals gain an advantage on federal contracts without changing anything about their technology.
The timing is awkward for Anthropic, which has been laying groundwork for a public offering. Watch for a petition for rehearing before the full D.C. Circuit or an appeal to the Supreme Court, and watch whether the Defense Department applies the same reasoning to other vendors.