D.C. Circuit Upholds Pentagon's 'Supply Chain Risk' Exclusion of Claude: A Court Rules Anthropic's Safety Refusals Are Themselves the Risk
Update (September 30–October 2, 2026): while the Pentagon fight drags on, Anthropic is expanding everywhere else in government. Anthropic moved Claude for Government to general availability for federal and state agencies on September 30, ending the public beta that began in July. Per the announcement: the platform delivers "Claude's coding and agentic work capabilities through a FedRAMP High authorized environment," with no seat fees — agencies pay for usage in fixed increments under a hard not-to-exceed cap — plus department-level spend allocation, audit logs built for ATO/IG requests, two-person approval on sensitive Anthropic-side operations, and local-only conversation history. Claude Code CLI and Claude for Microsoft 365 are rolling out in early access in the same environment. The Decoder's framing: Anthropic is going around the Pentagon to civilian agencies "as its fight with the Pentagon drags on."
The underlying ruling stands: a federal appeals court upheld the Pentagon's exclusion of Claude from military systems under the "supply chain risk" regime — a penalty normally reserved for foreign adversaries — on the theory that Anthropic's usage restrictions (its red lines on autonomous weapons and mass surveillance) are themselves a reliability risk. The February backstory, per ABC News: Hegseth told Amodei the Pentagon would invoke the Defense Production Act to force compliance if Anthropic didn't agree to its terms, and the WSJ reported Claude was used in the operation capturing Nicolás Maduro — which Anthropic declined to confirm. Meanwhile the Pentagon's GenAI.mil platform (The Pentagon's Frontier-Model Surge: 1.7M Users, 500K Power Users — and Grok and ChatGPT Heading to Classified Networks) is bringing Grok and ChatGPT to classified networks.
Why it matters: Anthropic is now running a two-track government strategy — the fastest-growing civilian-government franchise (FedRAMP High GA, procurement-ready terms) while litigating its right to say "no" to military use-cases. The D.C. Circuit's logic — that safety refusals are a supply-chain risk — is the sharpest legal statement yet of the government's side of the usage-policy fight, and it puts a price tag on the very caution the labs' safety brands are built on. The FTC probe and the White House accord both assume the labs control their models' deployment; the Pentagon case shows what happens when the customer with the biggest budget rejects that control.