US Military Officials Confer with Anthropic to Debate Claude Safety Measures

US military leaders, including Pete Hegseth, the Secretary of Defense, convened with executives from the AI firm Anthropic earlier this week to address ongoing tensions regarding government utilization of their advanced AI model. Hegseth has set a deadline of Friday for CEO Dario Amodei to comply with the Defense Department’s stipulations, warning of potential penalties should an agreement not be reached, as reported by Axios.

Anthropic, which brands itself as a frontrunner in AI safety, has been embroiled in a contentious dialogue with the Pentagon concerning permissible military applications of its large language model, Claude. While US defense officials advocate for broad access to Claude’s functionalities, Anthropic has reportedly resisted allowing its technology to be employed for mass surveillance or in autonomous weapons systems capable of making lethal decisions without human oversight. Despite the Department of Defense (DoD) integrating Claude into its operations, there have been threats to discontinue the collaboration over perceived obstacles presented by Anthropic.

The outcome of these discussions holds significance not only for Anthropic but also for the larger AI sector, as the balance of power between technology companies and governmental authority comes into play. The military use of AI products has long been a contentious issue, generating significant debate among AI researchers and ethicists regarding the implications of such applications. If Anthropic does not comply, defense officials have suggested punitive actions, including terminating a substantial contract with the company and categorizing it as a potential “supply chain risk,” as indicated by Axios.

Last July, the DoD finalized contracts with several prominent AI firms, including Anthropic, Google, and OpenAI, amounting to potential funding of up to $200 million. However, until the recent events, Anthropic’s Claude was the sole AI model authorized for use in military classified systems. Notably, the DoD recently signed a deal to permit the integration of Elon Musk’s xAI chatbot into classified operations, a product that has faced backlash for generating inappropriate content.

Counter to Anthropic’s stance, both xAI and OpenAI have accommodated the government’s requirements regarding the applications of their respective AI models. According to the Washington Post, a defense representative confirmed that OpenAI’s model has been approved for use in “all lawful purposes.” OpenAI did not immediately respond to requests for comments regarding their agreement with the Pentagon.

The meeting between Anthropic and military officials occurs against the backdrop of reports indicating that the US military utilized Claude during operations aimed at capturing Venezuelan leader Nicolás Maduro. Throughout the Trump administration, there has been a notable push for incorporating AI into military strategy, with the former president asserting that the US is committed to leading the global AI arms race.

Emil Michael, the Pentagon’s Chief Technology Officer and a former executive at Uber, has been vocal about the need for Anthropic to “cross the Rubicon” and acquiesce to the Defense Department’s terms. Michael articulated his opinion on that if a company seeks financial gain from government contracts, they should adapt their guidelines to accommodate lawful military uses, stating, “so long as they’re lawful,” in an interview with Defense Scoop last week.

Dario Amodei, Anthropic’s CEO, has consistently advocated for more comprehensive regulations on artificial intelligence. His firm has funded a political action committee that promotes stronger safeguards within the industry. Amodei’s political stances have included opposition to Trump during the 2024 presidential elections, and Anthropic’s recruitment of former Biden administration staff has reportedly caused tensions with pro-Trump investment groups, leading to a withdrawal of support from a venture capital firm previously interested in investing in Anthropic.

In recent years, the Pentagon has invested billions into developing AI-driven technologies, including unmanned aerial systems and automated targeting capabilities. This rapid advancement raises pressing ethical questions, particularly regarding how much authority should be granted to AI in making life-and-death decisions. The ongoing conflict in Ukraine highlights these concerns, as semiautonomous drones have been deployed in combat scenarios, operating with minimal human intervention.