OpenAI negotiating Pentagon deal that would include its safety "red lines"
Feb 28th 2026
Sam Altman told employees a potential Pentagon agreement would let OpenAI keep control of technical safeguards and named limits on military uses of its models, but the contract is not final.
- Sam Altman told staff a potential agreement with the Department of Defense is emerging but no contract has been signed.
- The deal would let OpenAI build and control its own safety stack of technical, policy, and human safeguards.
- OpenAI would retain control over which models are deployed and limit deployments to cloud environments rather than edge systems.
- The Pentagon reportedly agreed to include OpenAI's named red lines, such as bans on autonomous weapons, domestic mass surveillance, and critical decision making.
- The talks accelerated after a public fallout between the Pentagon and Anthropic that led to federal agencies being ordered to stop using Anthropic technology.
- Company leaders said foreign surveillance and risks to democracy are central concerns in the negotiations.