Anthropic is not done fighting the Pentagon. On October 13, 2026, the AI company behind the Claude models asked the full US Court of Appeals for the D.C. Circuit to rehear a case it lost in September.
The ruling it wants revisited lets the Defense Department keep treating a domestic AI developer as a national security risk.
What Anthropic is asking for
The filing is a petition for rehearing en banc. That is the legal term for asking every active judge on the court to weigh in, rather than leaving the final word to the panel that heard the case.
That panel ruled on September 25. It sided with the Pentagon, which had labeled Anthropic a “supply chain risk” under the Federal Acquisition Supply Chain Security Act, known as FASCSA.
The designation landed around February or March 2026. Its practical effect is sweeping: Pentagon personnel and contractors are barred from using Claude.
The government’s stated concern centers on Anthropic’s own contract terms. The company prohibits use of its technology for lethal autonomous warfare and for domestic surveillance.
The panel majority found “ample support” for the designation and pointed to the possibility that Anthropic’s model restrictions could constrain military operations.
Judge Henderson dissented from the majority’s decision.
Anthropic’s petition argues the panel got the law wrong. According to the filing, the court moved ahead without properly grappling with the due process and First Amendment questions raised by the Pentagon’s actions.
How a contract dispute became a security designation
The fight traces back to early 2026. Anthropic and the Pentagon were negotiating over military applications of the company’s AI, and those talks stalled.
When the discussions collapsed, the Trump administration escalated. It labeled Anthropic a national security risk, a significant and unprecedented move against a domestic AI firm.
Anthropic has also been fighting on a second front. In August 2026, Judge Rita Lin of a California district court ruled that related Pentagon actions amounted to unlawful retaliation against Anthropic’s public positions on AI safety.
That decision, issued August 27, characterized the government’s conduct as unconstitutional retaliatory measures. Its substantive core has so far gone unchallenged.
So the company now holds a win in California and a loss in Washington. The two courts examined connected government conduct and reached rather different places.
What this means
For Anthropic, the immediate issue is access. As long as the designation stands, the Defense Department and its contractor network cannot use Claude.
The en banc petition is an attempt to reset that outcome without going straight to the Supreme Court. If the full D.C. Circuit agrees to rehear the case, the September panel ruling could be reconsidered from scratch.
The California ruling framed the government’s actions as punishment for protected speech, while the D.C. Circuit panel treated the matter as a question of national security deference. Anthropic’s petition leans directly on that gap, arguing that the panel skipped the First Amendment and due process issues.
The next step is the court’s decision on whether to grant the petition. If it does, the full D.C. Circuit will have to decide whether national security deference or constitutional protections should carry more weight. If it declines, the panel ruling stands in that circuit, leaving two federal courts with sharply different views of the same government campaign against an AI company.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.