Where this stands
March 2026: Defense Secretary Pete Hegseth directs the Pentagon to designate Anthropic a supply chain risk.
27 August 2026: Judge Rita F. Lin in California rules for Anthropic on broader federal restrictions. That ruling still stands.
25 September 2026: The D.C. Circuit upholds the Pentagon designation, 2-1.
Now: Anthropic is considering an en banc rehearing or a Supreme Court petition. The Pentagon can keep excluding Claude.
What the disagreement was actually about
The Pentagon wanted to use Claude for "all lawful uses." Anthropic objected to two of them: mass surveillance and fully autonomous weapons.
Negotiations broke down. Defense Secretary Pete Hegseth then directed the department to designate Anthropic a supply chain risk under the Federal Acquisition Supply Chain Security Act of 2018 — a statute written to keep compromised or hostile suppliers out of federal systems.
Anthropic characterises the designation as unlawful retaliation for refusing those two uses. The administration describes it as a national security measure. Both descriptions are of the same sequence of events, which is what makes the case hard.
What the court decided
On 25 September 2026, the US Court of Appeals for the D.C. Circuit ruled 2-1 to uphold the designation, rejecting Anthropic's challenge.
Judge Gregory Katsas wrote the majority opinion, joined by Judge Neomi Rao. Judge Karen LeCraft Henderson dissented, arguing that the government had read the statute too broadly.
The dissent is the part to watch. A supply chain risk statute exists to exclude suppliers who are compromised, infiltrated or controlled by an adversary. Henderson's objection is that it was not written to cover a supplier who simply declined to sell a particular use. If that reading is right, the designation is a contract dispute dressed in national security language — and a 2-1 split with a reasoned dissent is exactly the shape of case that gets looked at again.
Two courts, two answers
This is the detail most summaries drop. Anthropic has not simply lost.
- D.C. Circuit, 25 September: upholds the Pentagon's designation. Claude stays out of Defense Department systems and defence contract work.
- Judge Rita F. Lin, Northern District of California, 27 August: sided with Anthropic on the broader federal restrictions arising from the same dispute. That ruling still stands.
So the position today is split: the Pentagon can exclude Claude, while a separate ruling limits how far the rest of the federal government can go. That is an unstable arrangement, and instability is usually what carries a case upward.
What it means in practice
If you are a defence contractor: Claude is excluded from Department of Defense systems and from defence contract work. That is operative now, not pending.
If you are a commercial Anthropic customer: nothing changes. This is about federal procurement, not product availability. Claude, Claude Code and the API are unaffected.
If you are choosing a vendor on usage policy: this is the clearest evidence yet of what a usage policy costs when it is tested. Anthropic gave up federal defence revenue rather than permit two applications.
If you are watching the precedent: the question is whether a supply chain risk designation can be used against a supplier for declining a use. That answer will apply to every AI vendor, not only this one.
Why it matters beyond Anthropic
Every frontier lab publishes a usage policy that forbids some applications. Those policies have largely gone untested against a customer with the power to retaliate.
This case tests one. If a designation of this kind survives, the practical cost of refusing a government use case rises sharply for every vendor — and the incentive shifts toward quieter policies with more room in them. If it is overturned, usage policies get a measure of protection they have never actually had.
That is a bigger question than one company's defence revenue, and it is unresolved.
What happens next
Anthropic says it disagrees with the decision and is considering all options. Two are available: asking the full D.C. Circuit to rehear the case en banc, or petitioning the Supreme Court.
Neither is guaranteed to be taken up. A 2-1 panel decision with a reasoned dissent on a question of statutory scope is a reasonable candidate for review, but most petitions fail, and nothing has been filed publicly.
FAQ
When did the Pentagon designate Anthropic a supply chain risk?
March 2026, directed by Defense Secretary Pete Hegseth after negotiations over permitted uses of Claude broke down.
Why was Anthropic designated?
The Pentagon sought to use Claude for all lawful uses. Anthropic refused two: mass surveillance and fully autonomous weapons. Anthropic calls the designation retaliation for that refusal; the administration calls it a national security measure.
What did the D.C. Circuit rule?
On 25 September 2026 it upheld the designation 2-1. Judge Gregory Katsas wrote the majority, joined by Judge Neomi Rao. Judge Karen LeCraft Henderson dissented, saying the government read the statute too broadly.
Has Anthropic won anything?
Yes. A 27 August 2026 ruling by Judge Rita F. Lin in California sided with Anthropic on broader federal restrictions stemming from the same dispute, and that ruling still stands.
Does this affect my Claude subscription or API access?
No. The designation concerns federal procurement and Department of Defense systems. Commercial access to Claude, Claude Code and the API is unaffected.
Is it over?
No. Anthropic is weighing an en banc rehearing or a Supreme Court petition, and the conflicting California ruling remains in force.