OpenAI Strikes Pentagon Deal for Classified AI Use With Built-In Safety Limits

New updates have been reported about OpenAI.

OpenAI has reached an agreement with the U.S. Department of Defense that will allow its AI models to be deployed on the Pentagon’s classified networks, positioning the company as a key defense technology provider while a rival exits the field. CEO Sam Altman said the contract embeds explicit prohibitions on domestic mass surveillance and mandates human responsibility for the use of force, including autonomous weapons, aligning OpenAI’s safety policies with existing defense law and policy.

The deal comes as Anthropic fails to agree to similar terms and faces presidential and Pentagon directives curtailing federal and defense-related use of its products, effectively shifting future government AI demand toward OpenAI and other compliant vendors. Altman emphasized that OpenAI will build technical safeguards and deploy engineers alongside defense personnel to ensure appropriate model behavior and security, and he urged the Pentagon to extend identical terms to all AI providers to reduce legal and political escalation.

Internally, Altman told staff the government will allow OpenAI to control a proprietary “safety stack” that can block specific tasks, with no obligation to modify the model if it refuses a request deemed unsafe, preserving a critical lever over how its systems are used in national security contexts. The contract implies a new, potentially material revenue stream in classified and defense applications, while also exposing OpenAI to heightened regulatory, reputational, and geopolitical scrutiny as AI becomes more tightly integrated into U.S. military and intelligence operations.

More than 60 OpenAI employees had previously joined hundreds at another major tech company in backing Anthropic’s push for stricter limits on military use, highlighting internal concern over defense work that management will need to manage as these deployments scale. With U.S. tensions and military activity in the Middle East intensifying in parallel to these announcements, OpenAI’s role in defense AI will be closely watched by policymakers, contractors, and investors assessing how safety commitments hold up under real-world operational pressure.