What happened
On 25 September 2026, a divided panel of the US Court of Appeals for the D.C. Circuit (Judges Katsas and Rao majority, Judge Henderson dissenting) upheld the Pentagon's 'supply chain risk' designation of Anthropic, rejecting claims that the blacklisting of its Claude models was arbitrary, unauthorized or unconstitutional. The court held the designation was supported under the Supply Chain Security Act and that balancing competing risks rests with the President and Secretary of Defense. The mandate is delayed to allow Anthropic to seek panel or en banc rehearing or Supreme Court review, leaving the blacklisting in force.
Why it matters
This is a binding national-security/enforcement outcome that materially governs how a major AI lab's models may be accessed and used by the US government and defense industrial base, and it expands executive authority to designate AI providers as supply-chain risks. It signals that frontier labs refusing government terms can be locked out of public-sector markets and sets a precedent for export-control-style restrictions on AI model access.
Action needed
AI labs and defense-facing deployers should track whether Anthropic obtains rehearing or stays; organisations contracting with DOD should confirm whether Claude is permissible in their environments and plan contingencies, and monitor whether similar designations are extended to other labs.