Altman Admits Pentagon Deal Was 'Sloppy'
OpenAI CEO Sam Altman is walking back a controversial Pentagon deal, calling it “opportunistic and sloppy” and promising to add new clauses limiting domestic surveillance. In a candid all-hands, Altman also told staff that government priorities increasingly dictate the company's operational decisions, signaling a major shift in how AI giants navigate geopolitical pressures.
The controversy erupted after OpenAI secured the Pentagon contract just hours after its primary competitor, Anthropic, was designated a "supply chain risk" for refusing a similar deal. Anthropic had insisted on contractual prohibitions against using its AI for mass domestic surveillance and for autonomous weapons systems without human oversight—terms the Pentagon rejected. Initially, OpenAI's agreement required compliance only with existing laws, which critics argued could still permit broad surveillance programs similar to the NSA's PRISM. The backlash was swift, with nearly 900 employees from OpenAI and Google signing an open letter urging a refusal of the Pentagon's demands. This public and internal pressure led to a user exodus toward competitors like Anthropic's Claude, which briefly became the top free app in Apple's App Store. In response, OpenAI revised the contract to explicitly ban the intentional use of its AI for domestic surveillance of U.S. persons, including through commercially acquired personal data. The new terms also require a separate contract modification for intelligence agencies like the NSA to use the technology. This episode marks a significant policy shift for OpenAI, which in January 2024 removed an explicit ban on "military and warfare" applications from its usage policy, replacing it with a broader rule against using its services to cause harm. The move aligns the company more closely with Microsoft, its largest investor, which has long pursued defense contracts, and diverges from Google's 2018 withdrawal from the Pentagon's Project Maven after employee protests. Altman's statement that operational decisions are ultimately up to the government highlights a fundamental governance challenge for AI developers. By drawing a line between building the technology and controlling its use, OpenAI is attempting to navigate the lucrative defense sector while deflecting direct accountability for how its AI is deployed in military operations.
