OpenAI Discloses Defense Contract Red Lines and Its Exit Strategy

OpenAI recently made public the details of its agreement with the U.S. Department of Defense, emphasizing that its technology deployed on classified networks is subject to strict restrictions. The contract defines three red lines: the technology must not be used for large‑scale domestic surveillance in the United States, must not be used to command autonomous weapon systems, and must not be involved in high‑risk automated decision‑making.

Compared with a prior agreement with competitor Anthropic, OpenAI says this deal includes more comprehensive safety safeguards. OpenAI retains full control over its own safety system, the technology is deployed via the cloud, and only vetted OpenAI personnel are involved throughout. The contract also contains strict breach clauses that give OpenAI the right to terminate the partnership if the government violates the terms.

The background is complex: on Friday former President Trump ordered the government to stop working with Anthropic, and the Pentagon subsequently listed Anthropic as a supply‑chain risk. OpenAI’s statement argues that its competitor should not be labeled similarly and notes that it has communicated its stance to the government.

In the past year the Pentagon has signed single‑award contracts up to $200 million with several AI labs to retain flexibility for defense use. OpenAI’s terms illustrate how technology firms are using contractual constraints to balance commercial interests with ethical boundaries.

OpenAI says it does not expect the government to breach the agreement but is prepared to end the collaboration if necessary.