Pentagon threatens to cut off Anthropic in AI safeguards dispute, Axios reports

The Pentagon is considering ending its relationship with artificial intelligence company Anthropic over its insistence on keeping some restrictions on how the US military uses its models, Axios reported on Saturday, citing an administration official.

The Pentagon is pushing four AI companies to let the military use their tools for "all lawful purposes," including in areas of weapons development, intelligence collection and battlefield operations, but Anthropic has not agreed to those terms and the Pentagon is getting fed up after months of negotiations, according to the Axios report.

The other companies included OpenAI, Google and xAI.

An Anthropic spokesperson said the company had not discussed the use of its AI model Claude for specific operations with the Pentagon. The spokesperson said conversations with the US government so far had focused on a specific set of usage policy questions, including hard limits around fully autonomous weapons and mass domestic surveillance, none of which related to current operations.

The Pentagon did not immediately respond to Reuters' request for comment.

Anthropic's AI model Claude was used in the US military's operation to capture former Venezuelan President Nicolas Maduro, with Claude deployed via Anthropic's partnership with data firm Palantir, the Wall Street Journal reported on Friday.

Reuters reported on Wednesday that the Pentagon was pushing top AI companies including OpenAI and Anthropic to make their artificial intelligence tools available on classified networks without many of the standard restrictions that the companies apply to users.