Pentagon stops using Anthropic’s AI tools

The statement closes a phase-out that ran past its own deadline. Defence Secretary Pete Hegseth labelled Anthropic a supply chain risk in February and gave the Pentagon six months, to late August, to drop it. Anthropic called the move unlawful and is suing the administration, the BBC reports.

Former defence officials and contractors told the BBC that Claude, including the Mythos models, was still in use as recently as last week. They say it sat inside Maven Smart System, the Palantir-run platform the Pentagon uses to organise intelligence, and that it helped with research, analysis and operations against Iran.

That account clashes with what the department said in September. Emil Michael, the under secretary for research and engineering, told DefenseScoop on 11 September that about 90% of the work had moved and that the Maven work had moved months earlier. The Pentagon now holds AI contracts with Google, xAI and OpenAI, according to the BBC.

For buyers of AI tools, the case shows how long a model stays in use once it is built into a larger platform, even after an order to remove it.