Skip to content
NewsRegulation

D.C. Circuit Backs Pentagon's Anthropic Blacklist, Splitting With California Ruling

· by Pondero Newsdesk

The short version

The U.S. Court of Appeals for the D.C. Circuit ruled 2-1 on September 25, 2026 that the Pentagon had grounds to bar Anthropic from military contracts over its refusal to drop safety restrictions, conflicting with an August ruling from a California federal judge on a related designation.

D.C. Circuit Backs Pentagon's Anthropic Blacklist, Splitting With California Ruling

A federal appeals court ruled Friday that Anthropic's refusal to let the Pentagon deploy Claude for autonomous weapons and domestic surveillance justified excluding the company from military contracts, a decision that now conflicts directly with a California federal judge's ruling on a related designation last month.

What

The U.S. Court of Appeals for the D.C. Circuit ruled 2-1 on September 25, 2026 to uphold the Department of Defense's "supply chain risk" designation of Anthropic, rejecting the company's claims that the label was retaliatory and exceeded the government's statutory authority, according to ABC News. Circuit Judge Gregory Katsas wrote the majority opinion, joined by Judge Neomi Rao; Judge Karen LeCraft Henderson dissented, per Defense One.

The designation traces to a February 24, 2026 Pentagon ultimatum demanding Anthropic drop its contractual restrictions on using Claude for lethal autonomous weapons and mass surveillance of Americans, per Defense One. Anthropic refused. Defense Secretary Pete Hegseth signed the designation on February 27, and the Pentagon announced it March 3, barring department employees and contractors from using Anthropic's products for official business.

The majority found the Pentagon had "ample support" to conclude that Claude's built-in restrictions, combined with the unresolved contract dispute, created a genuine national-security risk, per Courthouse News Service. Katsas wrote that the Pentagon "raises the deeply sobering prospect of overly constrained AI models shutting down unexpectedly and thus causing important military operations to fail," according to Military Times. Henderson's dissent argued that enforcing a model's own safety restrictions should not count as the kind of supply chain risk the statute was written to police, per ABC News.

The ruling gives the Pentagon a court-tested precedent for treating an AI vendor's own safety guardrails, not any external threat, as grounds for a supply chain risk label. For any company selling models into federal or defense-adjacent work, that raises the cost of holding firm on autonomous-weapons or surveillance restrictions: refusing a government customer's requested use case can now be argued as a security risk rather than a business decision, and a federal appeals court just said that argument wins. The ruling does not force Anthropic to drop its restrictions, but it removes one of the legal tools the company tried to use to block the designation while it negotiates or appeals further.

Context and reactions

An Anthropic spokesperson said the company "respectfully disagrees with the court's decision" and is "considering all options, including further review," pointing to a separate ruling that went the other way, per ABC News. In August 2026, U.S. District Judge Rita Lin in the Northern District of California ruled that Anthropic prevailed on one of two statutory justifications the government used for a related designation, finding the government's stated rationale was in part an attempt to "make a public example" of the company rather than a legitimate security response, per ABC News. That ruling and Friday's D.C. Circuit decision now stand in direct tension over the same underlying question.

What to watch next

Anthropic said it is weighing a request for en banc review by the full D.C. Circuit or a further appeal. Whether the conflicting district and circuit court outcomes get resolved through more litigation, or whether the Pentagon and Anthropic instead reach a negotiated settlement, is the next thing to track.

Sources