OpenAI Gives Pentagon Access to AI

The agreement materialized after Anthropic, the only AI model deployed across some classified DoD systems, reached an impasse in its own military contract negotiations. Anthropic established firm "red lines," refusing to permit its AI model, Claude, to be used for mass surveillance of American citizens or for fully autonomous weapons systems that operate without human control. In response to the stalemate, the Trump administration directed all federal agencies to cease using Anthropic's technology within six months. U.S. Secretary of Defense Pete Hegseth went further, designating the AI startup as a "supply chain risk," a label typically reserved for foreign adversaries, effectively barring military contractors from working with the company. OpenAI's deal includes its own set of principles, prohibiting the use of its technology for mass domestic surveillance and requiring "human responsibility for the use of force" in weapons systems. The company claims its agreement has more safeguards than Anthropic's, adding a third restriction against using its AI for high-stakes automated decisions, such as social credit systems. The agreement allows the Pentagon to use OpenAI's models for all "lawful purposes," consistent with existing law and policy. To enforce its restrictions, OpenAI states the deployment will be cloud-only, preventing use on edge devices for autonomous weapons, and that cleared OpenAI personnel will be involved in overseeing the systems. This partnership marks a significant policy reversal for OpenAI. The company, which began as a nonprofit, previously had a blanket prohibition on providing its technology for military applications, a rule that was quietly removed from its usage policies in 2024. The tension between AI labs and the military is not new. In 2018, Google abandoned its contract for Project Maven, an initiative to analyze drone footage, following widespread employee protests and resignations over the ethical implications of the work.