Appeals Court Lets the Pentagon Designate Anthropic a Supply-Chain Risk

A divided D.C. Circuit panel refused to second-guess the Trump administration’s supply-chain risk designation of Anthropic. The majority said the department had ample support that continued Claude integration presented a covered national-security risk because Anthropic encodes restrictions the company wishes to enforce.

A San Francisco federal judge tossed a parallel label under a different statute; Friday’s ruling leaves the other designation in place, so Pentagon blocking can continue. Both tracks face years of appeals.

Anthropic has said customers hesitated after the designations; it still touts growing sales and a possible IPO later this year. DoD has pointed to alternatives including SpaceX’s Grok, Google’s Gemini, and OpenAI models.

Same ruling, second desk cut: this is now an IPO-season overhang, not only a procurement spat. Danielle Cohen says Anthropic remains confident and is weighing all options—broader D.C. Circuit review or the Supreme Court.

The majority rejected due-process and speech claims, casting the fight as standard contract refusal: the department excluded Anthropic for not assenting to a term it deemed essential, not for wanting AI regulation. That framing drains the politics out of a deeply political fight.

We’re aligned with labs that won’t green-light autonomous weapons or domestic surveillance on current models. We’re watching the competitive effect: Google and OpenAI taking military deals Anthropic refused, while some of their own employees object.

If the designation survives into a public listing, every S-1 risk factor about government revenue gets louder. Useful AI for the state should not require labs to abandon hard limits. A court just said the Pentagon can punish them for keeping those limits anyway.

Paresh Dave, WIRED, September 25, 2026. The same panel had declined a temporary block in April. Secretary Pete Hegseth had called Anthropic’s stance a significant national-security risk.