According to a report by the BBC on October 5th, officials from the U.S. Department of Defense said in a statement on Monday that the Pentagon has ‘stopped using Anthropic’s products’. An Anthropic spokesperson declined to comment when asked about this statement.
Several former U.S. defense officials and contractors who have worked closely with the Pentagon on artificial intelligence (AI) projects told the BBC that last week, Anthropic’s Claude model was still used by the U.S. Department of Defense for research, analysis, intelligence collection, and military operations against Iran. These sources said that Claude is considered a key tool for filtering through vast amounts of information.
According to reports, since 2024, Anthropic has been used by the US government and military. It is also the first advanced AI company whose tools have been deployed by government agencies that handle confidential work. However, earlier this year, the Pentagon pressured the company to remove security measures for Claude and allow the military to use its AI products without restrictions.
After Anthropic refused on the grounds of "concerns about large-scale surveillance and autonomous weapons," U.S. Defense Secretary Pete Hegseth identified the company as a "national security supply chain risk" in February, and announced that the Pentagon would cease using its products by the end of August.
However, the deactivation did not complete as scheduled. The BBC stated that it is still unclear why there was a delay, but they mentioned that removing Claude from the Maven system, which is part of the Pentagon’s core intelligence processing platform and operated by Palantir, was complex and difficult, requiring several months of technical work to complete.
George City University's Security and Emerging Technologies Center senior research analyst Lauren Kahn said, "It is a fact that the Pentagon has not removed Claude's use from the entire system until now, which indicates that these devices are not plug-and-play."
"Once they're integrated into the system, removing them could be quite painful." said Kane, who had served as a US defense official.
A source familiar with the matter revealed that during the entire period of controversy with Anthropic, Pentagon employees used Anthropic’s Claude tool, including its Mythos model, which raised security concerns.
Sources familiar with the usage of this technology said that analysts and other defense officials relied on Maven to gather and organize visual data, including satellite images and drone footage. This data was then input into Claude and other large language models for tasks such as identifying potential military targets.
According to sources who visited Maven technology demonstrations earlier this year, officials often used this technology when preparing page-by-page reports and submitted these reports level by level. Some allies of the United States also utilized Maven to share intelligence with the Trump administration.
A former senior U.S. defense official said that when Hgeseth initially ordered the gradual elimination of Anthropic from government programs, “there were concerns that the government would switch from Anthropic to OpenAI and other tools.”
