OpenAI Inks Classified Military AI Deal
Separate from its more public Pentagon contract, OpenAI has reached a classified agreement to deploy its models on sensitive government networks. The deal is with the U.S. Department of Defense Operations and Warfare (DOW) and includes explicit guardrails for using the AI in mission-critical domains.
This agreement follows the collapse of similar negotiations between the Pentagon and rival AI lab Anthropic. Anthropic withdrew from the talks due to concerns over the use of its technology for mass surveillance and in fully autonomous weapons, refusing to agree to terms that would permit "any lawful use." In the wake of Anthropic's withdrawal, former President Donald Trump directed federal agencies to cease using Anthropic's technology, labeling the company a "supply chain risk." OpenAI secured its own deal with the Pentagon shortly thereafter. OpenAI has stated its agreement includes "red lines" that prohibit the use of its technology for mass domestic surveillance, to direct autonomous weapons systems, or for high-stakes automated decisions. The company says these guardrails are protected through a multi-layered approach including cloud-based deployment and oversight by cleared OpenAI personnel. The deal is part of the Pentagon's broader strategy to accelerate AI adoption for "enduring decision advantage." This initiative builds on previous efforts like Project Maven, a program started in 2017 to use AI for analyzing drone surveillance footage. OpenAI CEO Sam Altman admitted the announcement of the deal looked "opportunistic and sloppy" and was rushed. The agreement sparked backlash from some of OpenAI's own employees, nearly 100 of whom signed an open letter calling on the company to refuse demands for use in surveillance and autonomous killing. The contract is part of a larger Pentagon investment in AI, with agreements worth up to $200 million each signed with major AI labs like Anthropic, OpenAI, and Google over the past year. This partnership places OpenAI's technology within the Department of Defense's "AI-first" warfighting strategy, which aims to leverage advantages in data and AI models to innovate at "wartime speed." The agreement specifies that the AI system will not be used to "independently direct autonomous weapons" where human control is required by law or policy. However, the full contract has not been publicly released, fueling ongoing debate about the enforceability of these ethical guardrails.
