Pentagon Confirms End to Anthropic AI Tool Use After Security Designation
The Department of Defense has ceased using Anthropic's AI products months after labeling the company a supply chain risk.
The U.S. Department of Defense has officially stopped using artificial intelligence tools developed by Anthropic, according to an official statement provided to the BBC. This announcement comes months after Secretary of Defense Pete Hegseth designated the company a supply chain risk on national security grounds in February, with a planned cessation of use by late August.
Despite the directive, multiple individuals familiar with the matter indicated that Anthropic's Claude AI model was still in use by the department as recently as the week prior to the report. These sources, including former defense officials and contractors involved with Pentagon AI initiatives, described Claude as a critical tool for processing large volumes of information, aiding in research, analysis, intelligence gathering, and military operations. It was reportedly embedded within the Pentagon's Maven Smart System, a data platform operated by Palantir used for organizing intelligence and other data.
The integration of Claude into systems like Maven, which is deployed across the entire department and used by some U.S. allies for intelligence sharing, made its removal complex. Lauren Kahn, a senior research analyst at Georgetown's Center for Security and Emerging Technology and a former U.S. defense official, noted that deeply integrated AI tools are not simply plug-and-play, and their removal can be a difficult process.
Anthropic's AI tools had been utilized by the U.S. government and military since early 2024, marking the first instance of an advanced AI company's products being deployed for classified work within government agencies. However, earlier this year, the Pentagon reportedly pressed Anthropic to remove safety guardrails from its Claude model to grant the military unrestricted access. Anthropic declined this request, citing concerns about mass surveillance and autonomous weapons, which led to Hegseth's national security designation.
Anthropic responded to the designation, which is typically reserved for companies from countries posing a threat to the U.S., by calling it "unprecedented and unlawful" and initiated legal action to overturn it. As this legal battle progressed, the Pentagon entered into contracts with other major technology firms, including Google, xAI, and OpenAI, with OpenAI tools seeing increased adoption in some military departments.
The timeline of the Pentagon's order to phase out Anthropic tools, set for a six-month period starting February 27, coincided with significant geopolitical events, including U.S. and Israeli attacks on Iran. Sources indicated that Pentagon employees continued to use Anthropic's Claude models, including its Mythos variants, throughout the controversy, particularly for intelligence gathering and analysis within the Maven system. This system processes visual data like satellite imagery and drone footage, feeding it into large language models for tasks such as identifying potential military targets. Briefs for command chains frequently incorporated insights derived from this technology.
Meanwhile, Anthropic's relationship with the White House appears to be evolving. Anthropic CEO Dario Amodei met with former President Donald Trump twice at the White House in the week preceding the report. Following one of these meetings, Amodei acknowledged ongoing discussions regarding AI risks, stating that "if we work with the president, we can win safely." Trump, in turn, praised Amodei as "great" and "fantastic."