Pentagon drops Anthropic's Claude eight months after blacklisting AI firm

The US Department of Defense has halted all use of Anthropic's Claude AI, ending a relationship that collapsed after the company refused the military's demand to strip the model's safety guardrails.

According to the BBC, the Pentagon confirmed Monday (local time) that it had stopped using Claude and other Anthropic AI products. The move comes eight months after Defense Secretary Pete Hegseth designated Anthropic a "supply chain risk" company on national security grounds in February.

The Defense Department had originally planned to complete the phase-out by late August, but the actual removal took roughly two additional months. Claude had already been deeply embedded across core US military data systems, making a swift exit difficult.

Sources said the Pentagon continued to use Claude through last week for research, analysis and intelligence gathering. The model was also reportedly used to sort and analyze vast amounts of information during the Iran war.

Anthropic's AI was first introduced to the US government and military in 2024, quickly becoming one of the earliest advanced AI tools deployed within classified government operations.

The relationship soured earlier this year when the Defense Department called for Anthropic to remove Claude's safety restrictions and allow the military unrestricted use of the AI. Anthropic refused, citing concerns that the technology could be used for mass surveillance and autonomous weapons systems.

Hegseth subsequently designated Anthropic a supply chain risk. Anthropic filed suit against the Trump administration, calling the designation "unprecedented and unlawful," but the US Court of Appeals for the District of Columbia Circuit last month rejected the company's request to invalidate the blacklisting.

Removing Claude from active defense systems proved equally complicated. The model had been integrated into Maven Smart System, the Pentagon's large-scale data platform operated by Palantir. Maven is a core platform that consolidates and categorizes intelligence and data to support military analysis and decision-making.

Lauren Kahn, a senior research analyst at Georgetown University's Center for Security and Emerging Technology, said Claude's deep integration into Maven would have made rapid removal difficult.

"The fact that it has taken this long for the Defense Department to remove Claude from across its systems shows that this kind of technology is not something you can simply plug in and pull out," Kahn said.

Rival AI companies are moving in to fill the gap. The Pentagon signed a series of contracts with major AI firms — including OpenAI, Google and xAI — while the legal dispute was ongoing. In recent months, OpenAI's tools in particular have seen broader adoption within the US military.