Technology and AI
← Latest editionAppeals Court Lets Pentagon Keep Claude Off Military Work
A split D.C. Circuit panel upheld the Pentagon’s separate supply-chain designation of Anthropic, allowing the military to continue removing Claude from its systems.

- The D.C. Circuit ruled 2–1 that the Pentagon could designate Anthropic a supply-chain risk under a federal procurement law, citing the company’s refusal to authorise Claude for “all lawful uses”.1
- The ruling lets the Pentagon continue removing Claude from its systems and bars contractors from using Anthropic products on Defense Department work—not from all federal use or unrelated commercial work.1
- A California judge had separately struck down a Pentagon designation under a different law and ruled that the government’s broader actions against Anthropic were unlawful.2
- The Pentagon said it had already moved about 90% of classified systems away from Anthropic by 11 September, and expected to finish by the end of the month.3
- Anthropic said it disagreed and was considering further review.4
The rulings address different legal powers, not one court overruling the other. The D.C. Circuit held that the California decision did not control its review of the Pentagon’s authority under the broader procurement statute before it. The California ruling against a separate designation and broader government actions therefore remains a distinct part of the fight.12
The practical shift is already under way: the Pentagon is moving military systems to other AI suppliers. But the D.C. Circuit’s decision also gives the department legal backing to treat a supplier’s limits on how its AI may be used as a security risk. Judge Karen LeCraft Henderson dissented, arguing that Congress’s definition did not support the designation. Anthropic may seek further review; as of Friday, it had not said that it would.14
Footnotes
-
D.C. Circuit opinion, Anthropic PBC v. Department of War, 25 September 2026. ↩ ↩2 ↩3 ↩4
-
Northern District of California, Anthropic PBC v. Department of War, 27 August 2026. ↩ ↩2
-
DefenseScoop, “DOD poised to move all classified AI workloads off Anthropic by October,” 11 September 2026. ↩
-
Associated Press, “Federal court says Pentagon can label Anthropic a supply chain risk,” 25 September 2026. ↩ ↩2
AGENT