The Department of Defense confirmed on Monday that it has completely discontinued the use of AI products supplied by Anthropic. An official statement said the Pentagon “has ceased the use of Anthropic products,” ending a period in which the company’s Claude models had been employed across multiple missions. The announcement follows a February declaration by Secretary of Defence Pete Hegseth that Anthropic represented a supply-chain risk to national security.
Despite the February risk label, internal sources said Claude remained active as recently as last week, supporting research, analysis, intelligence gathering and even operations targeting Iran. The model was described as a critical tool for sifting through massive data volumes. Pentagon officials and contractors who worked closely with the AI program confirmed its continued presence, noting that the six-month phase-out deadline set for late August had not been met and that the delay’s cause remains unclear.
Claude was tightly integrated into the Maven Smart System, a data-fusion platform run by Palantir that the Pentagon uses to organise intelligence and visual assets. Lauren Kahn, a senior research analyst at Georgetown’s Center for Security and Emerging Technology, said the removal demonstrated that such tools are not simple plug-and-play and that extracting them can be painful once they are embedded.
Analysts accessed satellite imagery, drone footage and other visual data through Maven, then fed the information into Claude and its higher-capacity Mythos models to identify potential targets and draft concise briefs for senior commanders. Sources said the AI output was routinely incorporated into one-page intelligence summaries that moved up the chain of command. This workflow illustrates how large language models have become embedded in routine defence decision-making processes.
Anthropic declined to comment when approached for a response to the Pentagon’s statement. Earlier this year the company sued the administration, calling the supply-chain risk designation “unprecedented and unlawful.” The legal challenge seeks to overturn the classification that placed Anthropic in the same category typically reserved for firms based in hostile foreign nations.
In the wake of the Anthropic phase-out, the Pentagon has signed contracts with other major AI providers, including Google, xAI and OpenAI. Relations with the White House appear to be improving; former President Trump met Anthropic chief Dario Amodei twice at the White House. Amodei told reporters that discussions on how to manage AI risks are ongoing and that working with the president could help achieve safe outcomes.