Three more U.S. cabinet-level agencies — the State, Treasury and Health and Human Services departments — have stopped using AI products from Anthropic, aligning with the Pentagon’s shift toward rivals such as OpenAI under a new White House directive.

The widening federal pullback from Anthropic’s chatbot Claude marks a sharp rebuke of the San Francisco-based startup after contract talks with the U.S. Department of Defense broke down over AI guardrails and control of how the technology is deployed.

The Trump administration has clashed with Anthropic over restrictions on military and intelligence use of its AI, particularly around autonomous weapons targeting and domestic surveillance, sources said. OpenAI, backed by Microsoft and Amazon, announced a Defense Department deal last week, with CEO Sam Altman saying the company would clarify that its systems will not be intentionally used for domestic surveillance of U.S. persons.