Anthropic is not asking a new court for a new answer. Anthropic asks either the original three-judge panel to reconsider its ruling or the D.C. Circuit to rehear the case en banc, generally before its active judges. Filing the petition does not mean either form of rehearing has been granted.

The petition filed on 13 October targets the D.C. Circuit's judgment of 25 September. A panel voted 2–1 — Judge Katsas writing for the majority, joined by Judge Rao, with Judge Henderson dissenting — to uphold the Pentagon's decision to designate Anthropic a supply-chain risk under a 2018 federal acquisition law. The practical effect has not changed since March: the department cannot use Claude in its own work, and contractors cannot use Anthropic products in work performed for the department. A removal memorandum dated 6 March directed that the company's products be removed from department systems as soon as practicable, within 180 days.

The dispute underneath the label is narrow and well documented. Anthropic would not remove prohibitions in its terms covering mass surveillance of Americans and lethal autonomous weapons. The Pentagon wanted broader lawful use. When negotiations broke down, the designation followed. At points the department also raised concerns about Anthropic's ability to alter models already deployed; Anthropic argued it could not modify Claude after installation on classified systems, and the department later clarified that its determination did not depend on post-deployment control.

What Anthropic says the panel got wrong

Anthropic's core argument is about the reach of the statute. In its account, the law was "aimed at hostile foreign actors" and the panel read it far more broadly, in a way that could in principle reach ordinary government contractors whose disclosed contract terms the government dislikes. That distinction matters here because Anthropic's restrictions were not hidden. They were published terms, stated in advance, and they are the reason the company was designated.

The petition also raises process and speech. Anthropic says the designation was imposed without the pre-designation process the statute requires, and that the evidence points to retaliation for its public position on AI safety. The panel majority took a different view. The majority rejected Anthropic's retaliation and due-process challenges. It accepted the Pentagon's argument that Claude's embedded restrictions and Anthropic's refusal to permit all lawful military uses could create operational uncertainty, even without proof that Anthropic intended to sabotage military operations. The majority also distinguished Anthropic's contractual refusal from its public advocacy.

Read the dissent alongside the petition. Henderson's dissent interpreted the statute's reference to manipulating a product in light of surrounding terms such as sabotage and malicious interference. She argued that the provision requires intentionally subversive or deceptive conduct, rather than ordinary product design choices or openly disclosed contractual restrictions. Anthropic's rehearing request is, in large part, an invitation for the full court to adopt that narrower reading.

Two courts, opposite answers

This is not the only ruling on the Pentagon's campaign against Anthropic, and the other one went the other way. In August, a federal judge in California held a parallel designation, issued under a different Defense Department statute, unlawful, and treated Anthropic's speech as a substantial motivation for it.

That split is the real news value of a procedural filing. The D.C. Circuit's September ruling continues to uphold the FASCSA exclusion while Anthropic seeks rehearing. The separate California judgment concerned a different statutory designation and did not itself overturn the D.C. Circuit's decision. Granting rehearing alone would not establish that Anthropic ultimately wins the dispute. What the petition does is force the contradiction into one docket: one federal court has said a parallel label was unlawful retaliation, another has upheld this label, on materially the same underlying dispute.

For AI vendors selling to the federal government, the dispute tests whether disclosed model restrictions and their possible operational effects can support a national-security supply-chain exclusion. The majority accepted that possibility under FASCSA; Henderson would require conduct closer to intentionally subversive interference. The broader consequences for other procurement decisions remain a risk, not an automatic legal result of every use restriction. Which of those contracting environments applies is now up to the D.C. Circuit's willingness to rehear a case its panel has already decided once.

Sources

  1. [1] Anthropic, "Petition for Panel Rehearing or Rehearing En Banc", filed 13 October 2026 in Anthropic PBC v. U.S. Department of War, D.C. Circuit Nos. 26-1049 and 26-1162 (Document #2197902) — primary court filing, via the CourtListener docket for the caseRead source
  2. [2] U.S. Court of Appeals for the D.C. Circuit, opinion and judgment in Anthropic PBC v. U.S. Department of War, decided 25 September 2026 (Katsas, J.; Henderson, J., dissenting) — primary court document, via CourtListenerRead source
  3. [3] Reuters, reported 13 October 2026, syndicated copy via WNCY — wire report for the filing, the 2–1 September ruling, the parallel California ruling in August, and Anthropic's own billions-in-lost-business claimRead source
  4. [4] Unite.AI, "Anthropic Asks D.C. Circuit to Rehear Pentagon Supply-Chain Risk Case", 13 October 2026 — detailed summary of the petition, with a link to the filing hosted on CourtListener; used for the Rule 40 posture, the 25 September judgment date, the panel composition, the March designation timeline and the 27 August California rulingRead source