Pentagon threatens to cut off Anthropic in AI safeguards dispute: report

The Pentagon ‌is pondering ending its relationship with man made ​intelligence firm Anthropic over its insistence ⁠on conserving some restrictions on how the U.S. army makes use of its units, Axios reported on Saturday (February 14, 2026), citing ‌an administration professional.

The Pentagon is pushing four AI corporations to let the army ‌use their instruments for “all upright applications,” ‌including ⁠in areas of weapons pattern, intelligence series ⁠and battlefield operations, however Anthropic has now now not agreed to those phrases and the Pentagon is losing interest to demise after ​months of negotiations, according ‌to the Axios legend.

The different corporations integrated OpenAI, Google and xAI.

An Anthropic spokesperson acknowledged the firm had now now not discussed the use ‌of its AI model Claude for ​particular operations with the Pentagon. The spokesperson acknowledged conversations with the U.S. authorities ⁠to this level had centered on a selected plight of utilization policy questions, including laborious limits ‌round fully self sustaining weapons and mass home surveillance, none of which linked to latest operations.

The Pentagon didn’t at once acknowledge to Reuters’ quiz for comment.

Anthropic’s AI model Claude used to be feeble within the U.S. army’s operation to take hold of ‌feeble Venezuelan President Nicolas Maduro, with Claude deployed through ​Anthropic’s partnership with recordsdata company Palantir, the Wall Boulevard Journal reported on Friday (February 13, 2026).

Reuters reported ⁠on Wednesday (February 11, 2026) that the Pentagon used to be pushing high AI ⁠corporations including OpenAI and Anthropic to impress their man made intelligence instruments accessible on ‌classified networks without quite a lot of the comparable previous restrictions that the corporations apply to users.