With the clock ticking down on a strict military deadline, Anthropic has officially rejected the Department of Defense’s demands for unrestricted access to its artificial intelligence systems. Drawing a hard boundary against the Pentagon’s new contract terms, CEO Dario Amodei released a definitive statement late Thursday night.
Defense Secretary Pete Hegseth sparked this standoff earlier in the week by moving to renegotiate current military contracts across all major AI laboratories. Although competitors like OpenAI and xAI reportedly bowed to the new terms, Anthropic held firm, refusing to compromise its core safety principles regardless of the pressure.
Drawing the Red Line on Lethal AI
At the center of the dispute are two absolute restrictions Anthropic places on the military application of its Claude AI models. First, the technology cannot be used for mass domestic surveillance of Americans. Second, and perhaps more controversially for defense contractors, the company absolutely forbids deploying its systems in fully lethal autonomous weapons operating without human oversight.
Hegseth's High-Stakes Ultimatum
Tensions boiled over earlier this week when the White House summoned Amodei for a face-to-face meeting with Secretary Hegseth. Behind closed doors, Hegseth reportedly laid down the law: accept the revised terms by the close of business on Friday, or face the consequences.
Hegseth's ultimatum came with severe threats:
-
Instructing major defense contractors to immediately assess their operational dependence on Anthropic’s Claude models.
-
Threatening to officially designate the startup as a "supply chain risk"—a devastating classification traditionally reserved for hostile foreign entities and active national security threats.
-
Reportedly evaluating the use of the Defense Production Act to legally mandate Anthropic's compliance with military demands.
Pro-Military, Anti-Autonomous
Framing this refusal as a pacifist protest would be entirely inaccurate. Reaffirming his belief in the existential necessity of using AI to defeat autocratic adversaries, Amodei pointed out that his company proactively partners with the intelligence community and various military branches.
Going on the record, the CEO emphasized that his firm isn't interested in micromanaging specific military operations or limiting technology in an ad hoc fashion. Rather, the objection zeros in on a narrow, terrifying set of use cases where unchecked AI deployment directly undermines the democratic values the armed forces exist to protect.
Bracing for the Friday Deadline
Preparing for the very real possibility of retaliation as Friday's deadline arrives, Anthropic has drafted a comprehensive contingency plan. Should the Department of Defense sever ties today, the company guarantees full technical support to ensure a smooth transition to alternative AI providers.
To prevent any dangerous disruptions to active operations, critical intelligence missions, or ongoing military planning, Amodei confirmed that Claude will remain fully available to the armed forces under the originally agreed-upon terms for as long as the offboarding process requires.