US defense chief gives Anthropic Friday deadline to drop AI safeguards: Reports

By Michael Hernandez

WASHINGTON (AA) - US Defense Secretary Pete Hegseth gave Anthropic until Friday to agree to remove safety standards on its Claude AI model or face significant penalties, multiple reports published Tuesday suggest.

Hegseth met behind closed doors at the Pentagon earlier Tuesday with Anthropic CEO Dario Amodei, issuing his threat to either agree to the Defense Department's terms or have Anthropic listed as a "supply chain risk" or have the Defense Production Act invoked to force the company to comply, according to the reports.

Claude is currently the only AI model allowed to interact with the Pentagon's classified systems. Hegseth has been seeking to have Anthropic agree to new terms, which include using Claude for "all lawful purposes."

Anthropic, however, has insisted that its software not be used to conduct mass surveillance of Americans and that it not be used to develop autonomous weapons that can attack without human input.

Amodei reiterated his company's position during Tuesday's meeting, according to NPR.

The Defense Production Act allows a president to compel a private company to produce certain products during national emergencies. Being labeled a "supply chain risk" would set off a chain of negative consequences for Anthropic in which third parties would be forced to either do business with it or the Pentagon.

"The only reason we're still talking to these people is we need them and we need them now. The problem for these guys is they are that good," a defense official told the Axios news site ahead of Tuesday's meeting.

The Pentagon confirmed last week that its relationship with Anthropic "is being reviewed."

"Our nation requires that our partners be willing to help our warfighters win in any fight," Pentagon spokesperson Sean Parnell said in a statement provided to Anadolu.