Anthropic CEO Dario Amodei said Thursday that the company “cannot in good conscience” give the Pentagon unrestricted use of its model in all legitimate use cases, adding that the Pentagon’s threats will not change its position.
The artificial intelligence startup has been in tense negotiations with the Pentagon in recent weeks, and announced Thursday that those talks are still ongoing. Secretary of Defense Pete Hegseth has threatened to label Anthropic a “supply chain risk” or invoke the Defense Production Act to force the company to comply with the demands.
Anthropic wants assurances that its model will not be used for fully autonomous weapons or for large-scale domestic surveillance of American citizens, but the Pentagon wants to be able to use the model without these restrictions.
“It is the department’s prerogative to select the contractor that best aligns with its vision,” Amodei said in a statement. “But given the tremendous value that Anthropic’s technology brings to our nation’s military, we hope they reconsider.”
Mr. Hegseth met with Mr. Amodei at the Pentagon on Tuesday and gave Anthropic until Friday night to agree to the agency’s demands. A senior Pentagon official said the Pentagon sent Anthropic a “last and final proposal” on Wednesday night.
Pentagon chief spokesman Sean Parnell said Thursday that the Pentagon has “no interest” in using Anthropic’s model for fully autonomous weapons or conducting mass surveillance of Americans, noting that this would be illegal. He stressed that he hopes the company will agree to allow the agency to use its model for “all lawful purposes.”
“This is a simple, common sense request to prevent Anthropic from potentially endangering critical military operations and endangering our nation’s warfighters,” Parnell wrote in a post to X on Thursday. “We will not let any company dictate the terms of how we make business decisions.”
Anthropic won a $200 million contract with the Department of Defense in July, making it the first laboratory to integrate its model into mission workflows on classified networks.
The company’s rivals OpenAI, Google and xAI also won contract awards of up to $200 million from the Department of Defense last year. The companies agreed to allow the Pentagon to use their models for any lawful purpose within the military’s unclassified systems, but xAI also agreed this week to allow its models to be used in classified settings.
“Our strong desire is to continue to serve the Department of Defense and our warfighters with the two safeguards we have requested,” Amodei said. “If the Department chooses to offboard Anthropic, we will work to avoid disruption to ongoing military programs, operations, or other critical missions and allow for a smooth transition to another provider.”
Spotlight: Anthropic’s AI safety contradictions
