2-1 Court Backs Pentagon Blacklisting of Anthropic on AI Safety Guards

For defense-tech builders, the 2-1 appeals court decision cuts to the heart of a growing procurement dilemma: can a Silicon Valley AI company with ethical guardrails still be a trusted national security supplier? The majority sided with Defense Secretary Hegseth’s warning that "overly constrained AI models" could shut down mid-operation, effectively treating product-level safety constraints as a disqualifying supply chain risk. The ruling may force more AI defense suppliers to disclose exactly how their models will behave in autonomous and high-stakes military environments.

The U.S. Court of Appeals for the District of Columbia Circuit handed the Pentagon a significant victory on September 25, 2026, when a 2-1 panel upheld the Defense Department’s designation of Anthropic as a national security supply chain risk. The decision, authored by Judge Gregory Katsas, means the San Francisco-based AI startup remains barred from military contracts because it refuses to allow its products to be used for autonomous weapons or mass surveillance. The court majority rejected Anthropic’s argument that the designation was retaliation for the company’s stated views on AI safety and ethics, finding that the Pentagon acted reasonably in light of operational risks. Defense Secretary Pete Hegseth had argued that Anthropic’s safety restrictions could compromise military operations, and the court embraced that concern, writing of the "deeply sobering prospect of overly constrained AI models shutting down unexpectedly and thus causing important military operations to fail."

“

Court of Appeals for the District of Columbia Circuit handed the Pentagon a significant victory on September 25, 2026, when a 2-1 panel upheld the Defense Department’s designation of Anthropic as a national security supply chain risk.

The ruling marks a pivotal moment in an escalating legal and political battle that began in March 2026, when the Pentagon first designated Anthropic as a risk to national security supply chains. Anthropic has said the blacklisting has cost it billions of dollars in lost business and damaged its reputation just as it prepares for a highly anticipated initial public offering. The appeals court’s decision therefore carries consequences beyond government contracting: it may become a material risk factor for investors and a cautionary tale for AI companies deciding how to balance ethical guardrails with defense revenue.

Adding complexity, U.S. District Judge Rita Lin in San Francisco ruled last month in Anthropic’s favor in a parallel action brought under a different law, finding the administration unlawfully retaliated against Anthropic for its views on AI safety. Judge Lin also blocked the Trump administration’s government-wide ban on Anthropic and Hegseth’s order barring military contractors from doing business with the startup. The appellate ruling does not erase that district-level injunction, but it gives the government a strong precedential decision in the D.C. Circuit that may shape how other courts view the Pentagon’s autonomous-weapons and mass-surveillance restrictions. Anthropic said Friday it respectfully disagrees and is considering en banc review by the full appeals court.

From a regulatory perspective, the decision expands the executive branch’s discretion to treat a contractor’s product-safety policies—especially those limiting the deployment of AI in weapons systems—as a legitimate supply-chain risk. It also creates a federal legal landscape in which an AI company can win in one court and lose in another under different statutes, prolonging uncertainty for providers and procurement officials alike. Legal experts will likely focus on the interplay between the Administrative Procedure Act, First Amendment retaliation claims, and the national security deference the court afforded the Pentagon. The majority did not need to resolve whether Anthropic’s refusal to permit autonomous weapons use was expressive conduct protected by the First Amendment because it found the government had independent supply-chain reasons for the designation. That narrow path may invite further appeals.