U.S. Court Clears the Way for Claude AI to Be Removed From Military Systems

A U.S. federal appeals court on September 25 upheld the Pentagon’s decision to designate Anthropic as a supply-chain risk. The ruling was adopted by a vote of 2 to 1. The removal of Claude from the systems of military agencies and federal contractors may therefore continue. A majority of the judges considered the use of this artificial intelligence model a national security risk.

The Pentagon relied on 2 separate legal grounds in its actions against Anthropic. The company challenged each of them in different courts. A federal judge in San Francisco ruled 1 of the grounds unlawful in March, and that decision remained in effect last month. A court in Washington, however, did not overturn the Pentagon’s other decision. Anthropic may take the case to a broader panel of judges or the U.S. Supreme Court.

The dispute arose over the restrictions Anthropic placed on the use of Claude. The company did not agree to the model being used in fully autonomous weapons or to monitor people within the country. The court based its ruling on Anthropic’s refusal to accept a contractual condition that was important to the Pentagon. The judges also rejected the company’s claims that its rights to free speech and due process had been violated.

The parties signed a contract worth 200 million dollars (approximately 340 million AZN) in July 2025. The agreement was intended to test advanced artificial intelligence capabilities for U.S. national security. Reports published in February 2026 claimed that Claude had been used in the operation to detain Venezuelan President Nicolás Maduro. The disagreement between the parties over the purposes for which the model could be used then intensified.

Donald Trump instructed all federal agencies in February to stop using Anthropic technologies. In June, the company had to block federal agencies’ access to its most powerful models at the time. The Pentagon has not disclosed detailed information about how it replaced Claude with alternatives such as Grok, Gemini and GPT. Although some Google and OpenAI employees objected to military cooperation, the companies did not withdraw from the agreements.

Following the Pentagon’s decisions, Anthropic said some customers had become cautious because of their relationships with the federal government and that the company had lost revenue opportunities. No figures on the financial impact were subsequently disclosed. At the same time, the company says its sales have increased in recent months. Anthropic is preparing for a possible IPO, meaning the first public sale of its shares, by the end of 2026. The court battles could continue for years.