Anthropic won’t budge as Pentagon escalates AI dispute

Anthropic is currently under pressure from the Pentagon, facing a deadline to provide unrestricted access to its AI model for the U.S. military. As reported by Axios, Defense Secretary Pete Hegseth delivered a stern message to Anthropic CEO Dario Amodei during a meeting on Tuesday, indicating that failure to comply could lead to the company being classified as a 'supply chain risk.' This designation is typically reserved for foreign adversaries and could have severe implications for Anthropic. In the event of noncompliance, the Pentagon may invoke the Defense Production Act (DPA), which empowers the president to compel companies to prioritize production for national defense purposes. The DPA was previously activated during the COVID-19 pandemic to ensure companies produced essential medical equipment like ventilators and masks. Anthropic has consistently maintained its stance against the use of its technology for mass surveillance of citizens or for fully autonomous weaponry, demonstrating a firm resolve on these issues. Pentagon officials, however, argue that the military's technology use should adhere to U.S. law and constitutional limits, rather than the guidelines set by private companies. The potential use of the DPA in this context represents a significant shift in its application and highlights a growing trend of instability within the executive branch, as noted by Dean Ball, a senior fellow at the Foundation for American Innovation and a former senior policy advisor on AI during the Trump administration. He remarked that such actions could signal to companies that political disagreements may jeopardize their business. The situation is further complicated by ideological tensions within the administration, where figures like AI czar David Sacks have publicly critiqued Anthropic’s safety policies as overly progressive. Ball expressed concern that this ongoing conflict could deter investors and corporate managers from viewing the U.S. as a stable business environment, jeopardizing the foundational elements that have made the country a pivotal hub for global commerce. As the standoff continues, Anthropic appears resolute in its decision not to relax its usage restrictions. With reports indicating that Anthropic is the only frontier AI lab with access to classified Department of Defense (DOD) systems, the Pentagon's aggressive approach reflects its precarious position. The DOD currently lacks alternative options, although it has reportedly secured an agreement to use xAI's Grok in some classified systems. This lack of redundancy adds pressure to the Pentagon, which cannot afford to lose its relationship with Anthropic. Ball emphasized that if Anthropic were to terminate their contract, it would create significant challenges for the DOD, especially in light of a National Security Memorandum from the Biden administration that urges federal agencies to minimize reliance on a single classified AI system. The DOD's situation is precarious, relying on a single vendor without the ability to rectify this issue in the short term. TechCrunch has reached out to both Anthropic and the DOD for further comments.