Anthropic loses its Pentagon supply chain risk appeal, though judges accept it had no bad motive
A California court threw out the other designation because it needs bad intent, and this law doesn't.
On September 25, 2026, the federal appeals court in Washington, DC, turned down Anthropic's challenge to the Department of War's decision to cut Claude out of its supply chain. Secretary Pete Hegseth made that determination on March 3, 2026, under a 2018 procurement law, and two of the three judges held that he had the authority to do it.
Circuit Judge Katsas wrote the opinion, joined by Judge Rao. Judge Henderson dissented.
- Case
- No. 26-1049
- Decided
- September 25, 2026
- Opinion
- Circuit Judge Katsas
- Dissent
- Circuit Judge Henderson
- Removal
- within 180 days
Two lines in a contract
The fight was over two exceptions. Anthropic kept contract terms barring Claude's use for "lethal autonomous warfare" and "mass surveillance of Americans", and the Department wanted permission for "all lawful uses". Dario Amodei refused in writing on February 26, and he opened that statement by conceding the point the court would lean on.

Anthropic understands that the Department of War, not private companies, makes military decisions.
The court's reasons go beyond the contract, though. It lists occasions when Anthropic's restrictions got in the way: Claude refused queries from the CDC about research on preventing the spread of infectious disease, and an Anthropic executive "questioned the propriety" of a contractor's use of Claude in a military operation abroad. (The Department never named the operation. Anthropic's own filings included press reports tying it to the January 3 operation to capture Venezuela's president, Nicolás Maduro.)
Bad motive isn't required
What surprised me is how little the ruling holds against Anthropic. A federal court in California had already set aside a separate designation on August 27 under a different statute, one that requires an adversary and bad intent. The appeals court says it has "no quarrel" with the finding that Anthropic had no such motive, and rules against it anyway.
But as explained at length above, no such bad motive is required to support a designation under the much broader definition set forth in section 4713.
Henderson reads that law's words "or otherwise manipulate" as covering only deliberate, deceptive acts, like the sabotage listed beside them. Under the majority's reading, she writes, the next AI supplier that refuses a demand to loosen its policies faces the same choice Anthropic did.
a contractor’s honest and upfront enforcement of restrictions on a covered article’s use disfavored by the government
The majority's last word goes to who decides: "it is the President and the Secretary of War who must determine how best to balance the competing risks." Anthropic's newsroom had no post on the ruling when I checked on the evening of September 25.
More on Anthropic
- Anthropic's $11.6 billion Akamai cloud deal comes with a warrant for up to 5% of AkamaiSeptember 25, 2026
- “You've hit your session limit” in Claude Code (it used to say “Claude usage limit reached”)September 25, 2026
- “Your account does not have access to Claude. Please login again or contact your administrator.” and the other 403s in Claude CodeSeptember 25, 2026
- “API Error: Server is temporarily limiting requests (not your usage limit)” in Claude CodeSeptember 25, 2026