Anthropic holds firm against Pentagon on autonomous weapons and mass surveillance as deadline looms

Today at 5:01 PM ET, a Pentagon deadline expires for Anthropic to remove two safety guardrails from its Claude AI: prohibitions on mass domestic surveillance and fully autonomous weapons. Anthropic has publicly refused. CEO Dario Amodei stated the company "cannot in good conscience accede to their request" -- a direct rebuff of Defense Secretary Pete Hegseth, who threatened to terminate a government contract, designate Anthropic a "supply chain risk," and invoke the Defense Production Act to compel compliance.

The standoff marks a rare public split between a major AI company and the federal government over the terms of military AI deployment. Competitor xAI signed an "all lawful use" agreement with the Pentagon on February 23, granting Grok access to classified systems without restrictions. Google and OpenAI are reportedly in negotiations under similar terms, leaving Anthropic as the only major AI lab to refuse outright. Legal experts note that invoking the Defense Production Act -- a 1950 Korean War-era statute -- to force alterations to a commercial product's safety restrictions would be without precedent.