San Francisco-based AI company Anthropic, which launched as an ostensibly more ethically driven AI enterprise than competitors like OpenAI, appears to be hitting an impasse in its contract negotiations with the Pentagon.

Axios reports today that Anthropic has been butting heads with Pentagon officials as the company has been trying to place restrictions on how its technology may be used. Specifically, Anthropic has been seeking to limit how the military uses its Claude AI model, specifically when it comes to spying on American citizens.

The Pentagon — ahem, the Department of War — is seeking to have its use of Claude open to "all lawful purposes," as Bloomberg notes, and talks around this appear to be breaking down.

In an official statement to Axios, Pentagon spokesperson Sean Parnell says, "The Department of War's relationship with Anthropic is being reviewed. Our nation requires that our partners be willing to help our warfighters win in any fight. Ultimately, this is about our troops and the safety of the American people."

An Anthropic spokesperson gave a statement saying, "We are having productive conversations, in good faith, with DoW on how to continue that work and get these new and complex issues right."

Claude is reportedly already in wide use within the Pentagon, and was reportedly used in the early January mission to capture Venezuelan President Nicolas Maduro.

And an anonymous senior official at the Pentagon says that Defense Secretary Pete Hegseth is getting close to calling off the whole contract, regardless of how complicated that will be. "It will be an enormous pain in the ass to disentangle, and we are going to make sure they pay a price for forcing our hand like this," the official told Axios.

Axios notes that "existing mass surveillance law doesn't contemplate AI," and thus Anthropic feels it had better pump the breaks before it is suddenly at the center of a massive scandal or legal fight.

And this contract fight could become a threat to Anthropic's business overall, if the Pentagon deems the company a "supply chain risk."

Should that occur, all other companies who do business with the Pentagon will have to certify that they don't use Claude or any other Anthropic product — and as Axios notes, that could spell trouble for the AI company, as Claude is already in wide business use at some of the country's biggest companies.

This trouble for Anthropic comes just after an open resignation letter from a senior employee went viral in recent weeks, along with a separate viral post on X about how soon AI may be coming for the jobs of many Americans.

Top image: U.S. Secretary of War Pete Hegseth speaks during a meeting with Japanese Defense Minister Shinjiro Koizumi at the Pentagon on January 15, 2026 in Arlington, Virginia. Koizumi is on a visit to Washington for talks on regional security threats from China and North Korea and strengthening the Japan-U.S. security alliance with Hegseth. (Photo by Alex Wong/Getty Images)