Defense Secretary Pete Hegseth is threatening to blacklist artificial intelligence company Anthropic from working with the U.S. military over its refusal to relax safety standards.

The threat was made Tuesday during a meeting between Hegseth and Anthropic CEO Dario Amodei, according to two people with direct knowledge of the meeting who were not authorized to speak publicly.

Amodei was for several months. claimed that The use of AI in domestic mass surveillance and AI-controlled weapons is an ethical line that the company will not cross, calling such uses “illegal” and “vulnerable to abuse.” Mr. Amodeo reiterated this position on Tuesday, according to a person familiar with the Hegseth talks.

Hegseth said Anthropic would need to allow the US to use AI for any “lawful” purpose, which could include AI-driven warfare and surveillance.

Pentagon officials said the Pentagon plans to continue using Anthropic’s tools regardless of the company’s wishes.

If Anthropic does not back down, Mr. Hegseth plans to invoke the Defense Production Act, officials said. The law, typically invoked during national emergencies, forces companies to produce certain products deemed important to national security.

A senior Pentagon official said the move would force Anthropic to allow the military to use its tools “whether it wants to or not.”

The Trump administration’s retaliation against Anthropic is expected to include classifying the company as a “supply chain risk,” a move tantamount to putting Anthropic on a government blacklist, according to people familiar with the talks.

Anthropic’s hardline stance on domestic surveillance and AI weapons has earned it the label “woke AI” by Hegseth and other Trump administration officials. White House AI czar David Sachs helped write the draft presidential order Tech companies were targeted last year over this claim.

AI experts say “Woke AI” is a vague and ill-defined term that Trump officials appear to be using to describe all the safeguards of powerful AI tools and the belief that AI chatbots have liberal bias built into their models.

Rival AI companies such as OpenAI and Google have agreed to use their AI tools in all “legal” scenarios, as has Elon Musk’s xAI, which was approved for use in classified settings this week.

The Pentagon last summer awarded contracts worth up to $200 million each to Anthropic, Google, OpenAI, and xAI, but Anthropic was cleared for classified use for the first time after defense officials deemed it the most advanced and secure model for sensitive military applications.

The dispute with the White House comes at a time when Anthropic is under increased scrutiny as it plans to go public this year. It is unclear how the friction with the administration will affect investors.

Amodei noted that the company’s valuation and revenue have only increased since Anthropic clashed with Trump administration officials over how AI could be deployed on the battlefield.

“My biggest fear is that there will be too few ‘fingers on the button’ and essentially one or a few people will be able to operate a fleet of drones without the need for other humans to cooperate to carry out orders,” Amodei wrote. January essay Outlined his concerns. “I think we should approach fully autonomous weapons in particular with extreme caution and not rush to use them without proper safeguards.”