The U.S. Department of War is no longer using Anthropic's AI products, months after blacklisting the company, an official told the BBC.
The Pentagon "has ceased the use of Anthropic products," the department official told the public service broadcaster in a statement on Monday, two years after the U.S. government and military began using its tools in 2024.
Earlier this year, the Pentagon pushed for Anthropic to remove safety guardrails for the military's use, but the company refused, citing concerns of mass surveillance and autonomous weapons.
Secretary of War Pete Hegseth subsequently designated the firm a "national security supply-chain risk" in February, a term normally reserved for foreign companies it perceives as a threat to the U.S., pledging that the Pentagon would stop using its tools by late August. Anthropic sued the Trump administration following that move, calling it "unprecedented and unlawful." The Pentagon has since signed agreements with other major tech companies, including Google, SpaceXAI, and OpenAI.
Claude reportedly still used last week
However, Claude was reportedly still being used for intelligence analysis as well as military operations against Iran as recently as last week, the BBC said, citing multiple people familiar with the matter. The reason for the apparent delay is unclear, and the confirmed stop in use of Anthropic by the Pentagon has not previously been reported.
The people include former U.S. defence officials and contractors who have worked closely with the department on AI, according to the broadcaster, who said the Pentagon viewed Anthropic as a critical tool for sorting through vast amounts of information. It was incorporated into a larger data platform used by the Pentagon called Maven Smart System, operated by Palantir, for organizing intelligence and other data, including satellite imagery and drone footage, multiple sources told the BBC.
After Hegseth's decision to phase out Anthropic, there was some concern in the department about moving beyond the firm to OpenAI and others, a former senior U.S. defence official told the broadcaster. However, OpenAI's tools have been more widely adopted by some military departments in recent months, the BBC said, citing two people familiar with the department's AI use.
The Pentagon's move comes despite an apparent thawing in a previously frosty relationship between President Trump and Anthropic CEO Dario Amodei. Amodei, who has continued to warn about AI risks, told reporters at the White House last week that "if we work with the president, we can win safely," while Trump praised him as "great" and "fantastic" during the same briefing.
Anthropic declined to comment when reached by the BBC. The Latent has also reached out for comment.