Appeals court refuses to pause Pentagon’s supply-chain risk label on Anthropic

The ruling

A three-judge panel of the D.C. Circuit on Wednesday declined to temporarily block the Defense Department’s designation of Anthropic as a supply-chain risk, leaving in place restrictions that prevent Pentagon contractors from using the startup’s AI models on DOD contracts. The short, four‑page opinion argued the “equitable balance” favors the government, noting the court was balancing “a relatively contained risk of financial harm to a single private company” against how the Department of Defense secures vital AI during an active military conflict.

Anthropic had asked for an emergency injunction, saying the label already cost the company business and inflicted reputational damage that could ripple through its broader operations. The panel acknowledged Anthropic “will likely suffer some irreparable harm,” and said an expedited merits decision is warranted — but stopped short of pausing the designation.

The stakes and what's next

This decision comes after a different federal judge in California briefly blocked the Pentagon’s label, but it has been reported that a legal quirk required Anthropic to also seek relief at the D.C. Circuit. Supporters who celebrated the California win were warned not to pop the champagne yet — and the D.C. Circuit’s move makes clear why. An Anthropic spokesperson said the company is “grateful the court recognized these issues need to be resolved quickly” and remains confident the designations are unlawful; a DOD spokesperson did not immediately respond.

What happens next is high-stakes and messy. The clash traces back to Anthropic CEO Dario Amodei telling Defense Secretary Pete Hegseth he would not let the company’s Claude model be used for autonomous weapons or mass domestic surveillance, and it has been reported that Hegseth and former President Trump subsequently labeled Anthropic a supply‑chain risk. This is more than a regulatory wrinkle — it’s the latest front in the broader tug‑of‑war over who controls advanced AI and under what rules.

Sources: politico.com, Hacker News