Pentagon and Anthropic Clash Over AI Access: Friday Deadline Looms

The U.S. Department of Defense and artificial intelligence firm Anthropic are locked in a dispute over the terms of AI technology usage, with a Friday deadline set for a resolution. The Pentagon is seeking full access to Anthropic’s Claude model for “all lawful purposes,” while Anthropic is pushing for safeguards to prevent its AI from being used for mass surveillance or autonomous weapons systems.

The Core of the Dispute

The disagreement centers on Anthropic’s concerns about how its powerful AI model, Claude, could be utilized by the military. Anthropic has sought explicit restrictions to prevent the use of its technology in ways that conflict with its ethical guidelines. Specifically, the company wants to ensure Claude is not used for mass surveillance of American citizens or to power weapons systems that can produce targeting decisions without human intervention. CBS News reports that Anthropic CEO Dario Amodei reiterated these red lines, stating they have never impacted a military operation.

Pentagon’s Position and Concessions

The Pentagon maintains that its use of AI is already governed by existing laws and internal policies that prohibit the practices Anthropic is concerned about. Emil Michael, the Pentagon’s chief technology officer, stated the Defense Department has “made some very fine concessions,” including a written acknowledgment of federal laws restricting surveillance and existing Pentagon policies regarding autonomous weapons. However, Anthropic has deemed these concessions insufficient.

Defense Production Act and Potential Blacklisting

The Pentagon has threatened to invoke the Defense Production Act (DPA) to compel Anthropic to comply with its demands. Politico reports the deadline is 5:01 p.m. On Friday. The Pentagon has taken steps toward potentially blacklisting Anthropic, designating it as a “supply chain risk,” a move typically reserved for foreign companies like Huawei. The New York Post indicates Boeing and Lockheed Martin have been asked to detail their reliance on Anthropic’s Claude chatbot in preparation for a potential blacklisting.

Broader Implications

This dispute highlights the growing tension between the military’s need for advanced AI capabilities and the ethical concerns surrounding the technology’s development and deployment. The Pentagon is eager to leverage AI to counter advancements made by countries like China, as Michael emphasized, stating, “So we’ll never say that we’re not going to be able to defend ourselves in writing to a company.” Other AI companies, including Elon Musk’s xAI and Google, are also part of the Pentagon’s pilot program, with xAI reportedly on board with classified use.

The Friday Deadline

If Anthropic does not agree to the Pentagon’s terms by the 5:01 p.m. Friday deadline, the military plans to terminate its partnership with the company. This could have significant implications for both the Pentagon’s AI initiatives and Anthropic’s future collaborations with the U.S. Government.