Anthropic draws red lines as Hegseth sets Friday deadline to accept government terms

Anthropic has outlined its red lines in a meeting with the United States (U.S.) Secretary of War Pete Hegseth on Tuesday, February 24, 2026.

The government is pressuring the AI firm to provide unchecked access to its most powerful large language model, Claude, for military purposes and abandon its ethics rules.

According to the BBC, the U.S. secretary of war met Anthropic CEO Dario Amodei at the Pentagon and set a deadline until Friday for the AI giant to agree to the department’s terms or face penalties.

Hegseth threatened to remove Anthropic from the agency's supply chain if the demands were not met.

A source familiar with the discussion between Hegseth and Amodei has revealed the red lines shared by Anthropic CEO with the U.S. defence chief.

The limits set by the AI firm include:

• No AI systems making final military targeting decisions without human oversight.
• No use of Anthropic’s AI systems for mass domestic surveillance.

Another source said that Hegseth wants to invoke Defence Production Act on the company, in case it does not comply with his demands.

What is the U.S. Defence Production Act?

For context, the U.S. Defence Production Act (DPA) of 1950 is a federal law that grants the President, and by extension the Department of Defence (DoD), broad authority to force private companies into accepting government contracts for products, services, and materials essential to national defense, especially during emergencies.