The defense secretary’s desperation to get top Pentagon contractors to bend to his demands raised red flags among experts.
“It doesn’t make any sense,” Dean Ball, a former AI adviser to the Trump administration who helped shape the president’s AI plans, told Politico that Defense Secretary Pete Hegseth threatened Anthropic CEO Dario Amodei, 43, on Tuesday over the AI company’s contracts with the agency.
At a meeting between Hegseth and Amodei on Tuesday, Hegseth, 45, gave the CEO of the AI security and research company an ultimatum: Give the military unrestricted use of its AI models or face them being deemed a “supply chain risk” and cut off from the Pentagon.
Dario Amodei’s Anthropic has proven to be a Democratic island in a sea of Republican-friendly AI. /Anadolu, Getty Images
The meeting did not begin on a hopeful note. Before the meeting began, a defense official described it as “an unsuccessful meeting.”
Hegseth’s threats — to label Anthropic a “risk” and sever ties with the company, or to invoke the Defense Production Act (DPA), which requires the company to provide technology on Pentagon terms without current ethical restrictions — has experts scratching their heads on the matter.
“You’re telling everyone else that supplies to the Department of Defense that you can’t use Anthropic’s models, and at the same time you’re saying that the Department of Defense must use Anthropic’s models,” Ball said of Hegseth’s ultimatum to Anthropic, calling the move “incoherent.”
Hegseth, a former Fox News contributor, was criticized this month for posting a video of a workout with troops as tensions with Iran escalated. / Miguel J. Rodriguez Carrillo/AFP/Getty Images
Ball added that in proposing the two policies, “it’s a whole different level of madness to go up and say we’re going to do these two things.”
Amodai has so far resisted a push from the defense minister (who prefers to be called the “Secretary of War”) for unfettered, unrestricted access to Anthropic’s AI technology, specifically the large language model Claude.
Tensions between Anthropic and Hegseth have escalated amid reports that the Pentagon used Crowder to participate in the Trump administration’s invasion of Venezuela and the capture of Venezuelan President Nicolás Maduro earlier this month.
Axios reported that during Tuesday’s meeting, Hegseth referred to the Pentagon’s claim that Anthropic raised concerns with its partner Palantir about the use of Claude during the Venezuela mission, which Amodei denied.
Katie Sweeten, a tech lawyer and former Justice Department official, told Politico that Hegseth’s threats were “ambivalent.”
“I don’t know how you can use a DPA to take over this product while also saying it poses a huge national security risk,” Sweeten told Politico. He warned that Hegseth’s “very aggressive” negotiations could have devastating effects on future partnerships with Silicon Valley.
Experts warn that Hegseth’s negotiating tactics could cause problems for future contracts between the Pentagon and Silicon Valley. /Chip Somodevilla/Getty Images
The feud with Hegseth could derail a $200 million Pentagon contract with Anthropic signed in July to develop “artificial intelligence capabilities that advance U.S. national security.”
“The only reason we’re still talking to these guys is because we need them, and we need them now. The problem with these guys is they’re so good,” a Defense Department official told Axios about the AI company.
The Daily Beast has reached out to the Department of Defense for comment.
