Anthropic asks D.C. Circuit to rehear supply-chain risk case

Anthropic argues that the panel majority wrongly interpreted the supply-chain statute to cover model training before deployment.[1] It also argues that the designation was retaliation for its speech about AI safety.[1] Anthropic says the Secretary’s claim that the company had subversive intent was not credited by the majority.[1] The company contends that the ruling gives the Secretary virtually unchecked power over contractors, leaving them to choose between accepting the Secretary’s demands and risking a national-security designation.[1]

Anthropic says the Secretary bypassed statutory notice and an opportunity to respond before the designation.[1] It argues that the panel treated the lack of advance process as harmless on the ground that it would not have changed the Secretary’s decision.[1] The company says due-process precedents require notice and a chance to respond before a national-security designation unless there is a particularized need.[1] Anthropic also argues the panel’s reading expands a statute aimed at hostile foreign actors to presumptively cover every government contractor.[1] It says rehearing is warranted because the panel decided exceptionally important questions and conflicted with other court decisions.[1]

The Department of War notified Anthropic on March 4, 2026, that it had determined covered procurements presented a supply-chain risk.[3] The notice applied the designation to Anthropic products or services covered by the statute’s definitions.[3] Anthropic says Claude use is covered only when it is a direct part of a Department of War contract, and that the designation cannot restrict use or business relationships unrelated to a customer’s specific contracts.[4] The company says the vast majority of its customers were unaffected.[4] CNBC reported that the D.C. Circuit upheld the Pentagon’s decision in September.[2]

Anthropic said it would provide its models to the Department of War and national-security community at nominal cost, with engineering support, during the transition for as long as necessary and permitted.[4] Anthropic said its concerns were limited to exceptions for fully autonomous weapons and mass domestic surveillance, not military operational decisions.[4] It also said it supported frontline warfighters with applications including intelligence analysis, modeling and simulation, operational planning, and cyber operations.[4]