Pentagon official details the 'holy cow' moments that sparked rift with Anthropic

The Pentagon's chief of research and development has revealed the Department of Defense's deep concerns regarding Anthropic, particularly over the company's potential decision to cut off access to its AI technology during critical situations. In a recent appearance on the "All-In Podcast," Undersecretary of Defense for Research and Engineering Emil Michael recounted significant moments that led to the Pentagon categorizing Anthropic as a supply chain risk, effectively blacklisting the AI firm. Michael highlighted a notable instance where Anthropic's CEO, Dario Amodei, suggested that a simple phone call could resolve the ongoing issues concerning the deployment of the company’s AI models. This suggestion came during discussions about the Pentagon's critical missile defense initiative, referred to as the Golden Dome. Michael expressed his frustration, emphasizing the irrationality of having to make such a call during an urgent situation, questioning the reliability of Anthropic's AI in high-stakes scenarios. The conversation took a serious turn when Michael mentioned an unsettling incident following a targeted U.S. operation in Venezuela aimed at capturing President Nicolás Maduro. An unnamed executive from Anthropic reportedly inquired whether their AI had been involved in the operation, raising alarms within the Pentagon. Michael disclosed that this prompted significant concern over the potential ramifications should Anthropic's AI malfunction during future military engagements. As discussions intensified, Michael accused Anthropic of turning the dialogue into a public relations battle, particularly as Amodei voiced concerns regarding the Pentagon's terms and their implications for AI misuse. The tensions escalated to a point where Michael publicly labeled Amodei as a "liar" and accused him of having a "God-complex." In a historic move, the Pentagon recently informed Anthropic of its decision to designate the company as a supply chain risk, a rare and severe classification. In response, Amodei claimed that recent conversations with the Pentagon had been constructive, though Michael later stated that no talks were currently in progress. Anthropic has indicated plans to legally challenge this designation, especially since it could hinder defense contractors from engaging with the company. Michael clarified that this action was not punitive but rather a necessary precaution given Anthropic's potential biases in AI model outputs. He expressed his distrust in allowing companies like Lockheed Martin to utilize Anthropic's technology for military applications. Conversely, he acknowledged xAI and Elon Musk for their willingness to collaborate with the Pentagon under acceptable terms, alongside praising OpenAI for developing an alternative system capable of operating in classified environments, a move that aims to phase out reliance on Anthropic's AI. As the situation continues to unfold, the Pentagon’s approach to AI integration and its partnerships with tech companies will be closely watched, especially in light of the challenges posed by rapidly evolving AI technologies.