Pentagon may cut ties with Anthropic over AI usage limits

The Pentagon is considering ending its relationship with artificial intelligence company Anthropic because the firm has refused to lift certain restrictions on how its AI models can be used by the US military, Axios reported, citing a US administration official.

Defense officials are pushing several AI companies — including OpenAI, Google, xAI and Anthropic — to allow military use of their tools for “all lawful purposes,” such as weapons development, intelligence gathering and battlefield operations. Anthropic has resisted these demands, maintaining limits on applications like fully autonomous weapons and mass domestic surveillance, and Pentagon officials are reportedly growing frustrated after months of talks.

An Anthropic spokesperson said that discussions with the US government have focused on policy questions rather than specific operational deployments and that some usage restrictions — particularly on autonomous weapons and surveillance — remain in place.

The reported dispute comes amid broader debates about the role of advanced AI in national security, including efforts by the Pentagon to expand access to leading models on classified networks with fewer restrictions.

Anthropic’s AI model Claude has been involved in US military activities, including use via a partnership with data firm Palantir in a mission related to the capture of former Venezuelan President Nicolás Maduro, according to previous reporting by the Wall Street Journal — though Anthropic has said it has not discussed specific operations with the Pentagon.

The Pentagon had no immediate comment on the Axios report.