Pentagon stops using Anthropic AI tools months after blacklisting company, BBC told

What the report says

The US Department of Defence has ended its use of Anthropic’s AI tools, according to a Pentagon official cited by the BBC. The move follows months of tension between the company and defence officials after Defence Secretary Pete Hegseth designated Anthropic a national security supply-chain risk earlier this year and said the Pentagon would phase out its products.

BBC reporting says that, despite that deadline, people familiar with the matter believed Anthropic’s Claude model was still being used until recently in research, analysis, intelligence gathering and some military activity. The system was also described as part of the Pentagon’s Maven Smart System, a data platform run by Palantir that helps organise large volumes of defence information. According to the sources, Claude was used with other tools to process material such as satellite imagery and drone footage and to help prepare briefs.

Anthropic has previously said the government’s designation was “unprecedented and unlawful” and has challenged it in court. The company had been used by US agencies since 2024 and was among the first advanced AI providers deployed in classified government work, but the Pentagon later pushed for broader access to its tools. Anthropic refused, citing safety concerns, while the Trump administration has also signed contracts with other AI firms including Google, xAI and OpenAI.

The BBC says the latest Pentagon statement confirms the company is no longer being used, though the timing and practical process of removing AI systems from defence platforms appears to have been complex.