Pentagon Secures Multi-Hundred-Million AI Prototyping Deals with Major Tech Firms

A lawsuit filed under the Freedom of Information Act produced over 400 pages of contracts between the Department of Defense and AI firms Anthropic, Google, OpenAI and xAI, each valued at up to $200 million. The agreements task the companies with creating prototypes to enhance military advantage, covering applications from automated decision-making to logistics and intelligence analysis. All contracts contain an "information sharing" provision that obliges the firms to supply large-language models, benchmark data and strategic briefings, while receiving Pentagon input on AI adoption.

The firms will also deliver training seminars, conduct tabletop simulations, and provide risk-forecasting for future AI capabilities. A dispute arose when Anthropic refused to sign an amendment that would allow deployment on classified networks without restrictions on domestic surveillance and autonomous weapons, leading to a temporary ban that was later overturned by a judge. Subsequent amendments with the other firms proceeded, including clauses embedding OpenAI engineers within military units and omitting oversight sections.