Court upholds Pentagon's blacklisting of Anthropic

A US appeals court has ruled that the US Department of Defense, the Pentagon, could lawfully label the AI company Anthropic a "supply chain risk," meaning a security risk in the supply chain. This means Anthropic's chatbot Claude remains banned from use at the Pentagon.

The case dates back to earlier this year, when the Trump administration ordered government agencies and military partners to stop working with Anthropic. The background was that Anthropic would not give the Pentagon free rein to use the company's AI for "all lawful purposes." Anthropic has rules for what its AI may not be used for, including with regard to autonomous weapons and surveillance of US citizens.

The appeals court in Washington D.C. decided the case with two judges in favor and one against. The majority found that the Defense Department had sufficient documentation that continued use of Claude in its systems posed a real security risk. The dissenting judge disagreed.

The ruling is a setback for Anthropic, which in August won a similar case in a court in California. There, the company was partly successful, so other government agencies and contractors outside the Defense Department can still use Anthropic's products. The two conflicting rulings may end up having to be decided by the US Supreme Court.

An Anthropic spokesperson says the company "respectfully disagrees" with the ruling and is considering appealing further.

For regular users of Claude, the ruling changes nothing right now; the ban only applies to the Pentagon's own systems and the companies that work directly for the US military. The case is, however, yet another example of how large AI companies' ethical guidelines for the use of their technology can come into conflict with government demands for unrestricted access, especially for military purposes.