OpenAI announced on 28 February 2026 that it has entered into an agreement with the United States Department of War, marking a formal partnership between the AI provider and the military agency [source]. The blog post confirms the arrangement but provides limited detail on the scope of services, data-sharing protocols, or safeguards governing military use of OpenAI's models.

The announcement states that OpenAI will provide "tailored AI capabilities" to support departmental operations, though it does not specify which models or API endpoints are included. No information is given on whether military applications will use standard commercial models or modified versions with altered safety filters.

OpenAI's usage policies have historically prohibited use of its models for weapons development, military surveillance, and certain categories of harm. The blog post does not clarify whether these restrictions remain in force under the new agreement or have been modified to accommodate military requirements.

The Department of War, re-established in 2025 following legislative reorganisation of defence agencies, oversees combat operations and strategic planning. The partnership raises questions about model behaviour in high-stakes contexts where errors or hallucinations could have operational consequences.

OpenAI states that the agreement includes "ongoing dialogue on responsible deployment," but does not describe independent oversight mechanisms or third-party auditing. The post does not address whether military use cases will be subject to the same rate limits, content moderation, or logging practices as commercial deployments.

No technical documentation or updated terms of service have been published alongside the announcement. Users and researchers have no visibility into how OpenAI models may behave differently when deployed in military contexts compared to civilian applications.