The artificial intelligence company Anthropic is dropping its core safety pledge in scaling its AI models. The company’s reversal comes after Defense Secretary Pete Hegseth issued an ultimatum to Anthropic’s CEO, Dario Amodei, to grant the Pentagon full access to the company’s AI models by Friday. Hegseth reportedly threatened to cancel a $200 million Pentagon contract and to declare Anthropic a “supply chain risk” — and even said he might invoke the Defense Production Act to force the company to tailor its AI models for military use. Amodei had previously warned against the use of artificial intelligence to help the U.S. government conduct mass surveillance and develop autonomous weapons. This comes as NBC News reports that in December, Anthropic agreed to allow the U.S. government to use its AI systems for missile and cyber defense purposes.

Anthropic Drops Safety Pledge as Hegseth Demands Pentagon Access to AI Models

HeadlineFeb 26, 2026