Dario Amodei, the CEO, made his statement just under 24 hours prior to the deadline set in the Pentagon’s ultimatum.
Dario Amodei, the CEO, made his statement just under 24 hours prior to the deadline set in the Pentagon’s ultimatum.
Less than 24 hours prior to the deadline imposed by the Pentagon’s ultimatum, Anthropic has declined the Department of Defense’s request for unrestricted access to its AI.
This is the culmination of a heated exchange of public statements, social media discussions, and clandestine negotiations, pivoting on Defense Secretary Pete Hegseth’s wish to renegotiate existing contracts with AI labs linked to the military. Yet, Anthropic remains firm on its two established boundaries: no mass surveillance of U.S. citizens, and no lethal autonomous weaponry (or arms allowed to eliminate targets without human intervention). OpenAI and xAI have reportedly accepted the new stipulations, while Anthropic’s refusal led to CEO Dario Amodei being called to the White House this week for discussions with Hegseth, where the Secretary allegedly presented an ultimatum for the CEO to acquiesce by the end of the business day Friday or face consequences.
In a statement released late Thursday, Amodei stated, “I have a profound belief in the critical role of employing AI to protect the United States and other democratic nations, as well as to overcome our autocratic rivals. Accordingly, Anthropic has proactively endeavored to implement our models for the Department of War and the intelligence community.”
He noted that the company has “never objected to specific military missions or sought to restrict the utilization of our technology in an arbitrary way,” but in “a limited number of scenarios, we believe AI can weaken, rather than support, democratic principles” — specifically pointing to mass domestic surveillance and completely autonomous weaponry. (Amodei remarked that “partially autonomous weapons … are essential for safeguarding democracy” and that fully autonomous arms may sometime “become vital for our national defense,” emphasizing that “currently, cutting-edge AI systems are simply not dependable enough to operate fully autonomous weaponry.” He did not exclude the possibility of Anthropic submitting to military usage of fully autonomous weapons in the future but indicated they are not prepared at this time.)
The Pentagon had already reportedly urged major defense contractors to evaluate their reliance on Anthropic’s Claude, a move that could mark initial steps to label the firm a “supply chain risk” – a public warning the Pentagon has recently issued (a designation usually reserved for national security threats). Additionally, the Pentagon was reportedly contemplating invoking the Defense Production Act to ensure compliance from Anthropic.
In his statement, Amodei asserted that the Pentagon’s “threats do not modify our stance: we cannot, in good faith, comply with their demands.” He further added, “If the Department opts to sever ties with Anthropic, we will strive to ensure a seamless transfer to another provider, mitigating any disruption to current military strategies, operations, or essential missions. Our models will remain accessible under the broad terms we have proposed for as long as necessary.”
