OpenAI Inks New Classified Deal with Pentagon Wing
Separate from its more public contract, OpenAI has secured a classified agreement with the Department of Defense's Operations Wing (DOW). The deal will deploy OpenAI's models on secure government networks and includes new "guardrails" for ethical oversight, signaling a deeper integration of AI into sensitive national security operations.
This agreement follows the collapse of a similar deal between the Pentagon and rival AI firm Anthropic. Anthropic refused to remove safeguards prohibiting its AI from being used for mass surveillance or fully autonomous weapons, leading the Trump administration to designate the company a "supply chain risk" and order federal agencies to cease using its technology. OpenAI's deal materializes after the company quietly removed language from its usage policy in January 2024 that banned "military and warfare" applications. The previous policy explicitly forbade use in "weapons development," but the new terms focus on a broader injunction not to cause harm. In response to criticism, OpenAI and the Pentagon revised the initial agreement's broad "any lawful applications" language. The updated contract now explicitly states the AI systems cannot be "intentionally used for domestic surveillance of US persons and nationals," a key point of contention in the negotiations with Anthropic. OpenAI states it has three main "red lines" embedded in the contract: no mass domestic surveillance, no use for directing autonomous weapons systems, and no high-stakes automated decisions like social credit systems. The company says it will use cleared OpenAI personnel and cloud-only deployment to help enforce these limitations. The deal is part of the Pentagon's broader "AI-first" strategy, which aims to fundamentally re-imagine military operations with artificial intelligence at their core. A key procurement criterion for this strategy is ensuring the latest commercial AI models are available to military users within 30 days of their public release. While OpenAI has provided assurances, the lack of a binding federal AI law in the United States means the company is operating more on self-imposed policy than statutory protection. This has led to skepticism from critics who argue that early compromises on military contracts could set long-term precedents.
