Anthropic suffered a defeat in the United States judiciary as it tried to overturn the designation by the Trump administration that considers the artificial intelligence company a risk to the supply chain and to American national security.
A federal appeals court in Washington rejected, by a 2-1 vote, the company’s appeal on Friday. Still, a separate decision by a California federal judge that suspended the federal government’s prohibition on using Anthropic’s technology remains valid for now.
The decision represents a victory for the Trump administration, which has been waging a high-stakes dispute with Anthropic since negotiations failed over the use of the company’s AI models by the U.S. Armed Forces.
According to Judge Karen LeCraft Henderson, Anthropic refused to relax contractual clauses that prohibit the use of Claude in lethal autonomous weapons systems or for domestic surveillance.
According to the judge, the Department of Defense had sufficient grounds to conclude that the ongoing integration of Claude into government information systems and those of its contractors posed a risk to national security as provided by law.
At the same time, a decision issued in August by a federal judge in San Francisco remains in effect. In that case, the judge concluded that the prohibition on the government’s use of Anthropic’s technology had not been adequately justified, since the applicable law defines supply chain risk more narrowly.
Henderson stated that the Washington court did not challenge that conclusion, noting that Anthropic did not act in bad faith in its negotiations with the government. However, according to her, the characterization as a supply chain risk can be upheld by another, broader statute that does not require proof of malicious intent.
In a statement, Anthropic said it “respectfully disagrees” with the decision.
“Another federal court has already concluded that a parallel government designation was illegal. We remain confident in our position and are evaluating all options, including new appeals,” the company said.
Emil Michael, Undersecretary of Research and Engineering at the Department of Defense, celebrated the decision on social media.
“The hammer of justice crushed Anthropic’s arguments,” he wrote. According to him, American service members “will sleep more soundly knowing that no private company can insert its opinions into the chain of command.”
The dispute comes at a time of intense competition among American AI companies to roll out increasingly advanced models and prove to investors that the technology can generate profits.
In February, Anthropic sought assurances that its tools would not be used for mass surveillance of American citizens or for autonomous weapons systems. The government, however, wanted access to Claude without the restrictions imposed by the company.
Following the impasse, the Pentagon designated Anthropic and its products as a risk to the United States supply chain, directing federal agencies to review or withdraw guidelines tied to the company. Normally, this kind of designation is reserved for companies from countries considered adversaries by the United States.
Anthropic responded by suing the Department of Defense in both California and Washington, accusing the government of retaliating against the company after the negotiations failed.
The Washington case concerns specifically the national security risk designation, while the California case challenges the broader prohibition on using the company’s technology in government contracts.
