OpenAI details its Department of War agreement with explicit limits on surveillance and autonomous weapons

What OpenAI announced

On February 28, 2026, OpenAI published details of its agreement with the Department of War for deploying advanced AI systems in classified environments. The company then added an update on March 2, 2026 after further discussions with the Department. OpenAI argues that the arrangement creates a tighter framework than earlier classified AI deployments because deployment architecture, technical safeguards, contract language, and human oversight are all specified together.

The company says three red lines guide the work: no use of OpenAI technology for mass domestic surveillance, no use to direct autonomous weapons systems, and no use for high-stakes automated decisions such as social-credit-style systems. OpenAI also said it asked the Pentagon to make the pathway available to all AI companies rather than structuring the opportunity as a single-lab channel.

How the safeguards work

• Cloud-only deployment: OpenAI says the systems will run through a safety stack that OpenAI operates, rather than through stripped-down or guardrails-off models.
• No edge deployment: the company says it is not deploying models on edge devices, where they could create additional autonomous-weapons risk.
• Contractual limits: OpenAI quotes language stating the system will not independently direct autonomous weapons where law or policy requires human control and will not assume other high-stakes decisions that require human approval.
• Human involvement: cleared forward-deployed OpenAI engineers, along with safety and alignment researchers, remain in the loop.

The March 2 update made one point much more explicit. OpenAI said the Department agreed the tools will not be used for domestic surveillance of U.S. persons, including through commercially acquired personal or identifiable information. The update also states the services will not be used by Department of War intelligence agencies such as the NSA unless a new agreement is negotiated.

Why it matters

Frontier labs rarely publish this much operational detail about military AI contracts. OpenAI is effectively treating deployment architecture and compliance language as part of the product, not just as legal paperwork in the background.

That makes the announcement important beyond one Pentagon deal. It creates a public reference point for how future national-security AI deployments may be debated: not only whether labs should participate, but under what technical, contractual, and governance constraints they are willing to do so.

Source: OpenAI