Documents released through a Freedom of Information Act lawsuit reportedly show that the Pentagon sought an OpenAI system with fewer restrictions for national security work.
The document described models intended to operate with “minimal refusal rates” during certain military applications. It covered testing, evaluation, and improvement of OpenAI models for defense-related activities.
OpenAI and the Pentagon disputed that wording. OpenAI spokesperson Nate Evans said the phrase was not in the executed contract and described the released document as an earlier Department of Defense draft submitted before OpenAI provided feedback. A Department of Defense representative also said the phrase does not appear in any active contract.
A Pentagon lawyer initially confirmed the document before withdrawing that statement. Pentagon research official Trevor Tiedeman said there may have been confusion between released records and executed agreements and said he would examine how the document passed through review procedures.
The Intercept had requested completed contract records connected with military AI agreements. OpenAI has not publicly released the complete contract, and significant sections of the agreement remain redacted.
The updated military agreement allows OpenAI services across classified government networks. OpenAI has said the agreement restricts autonomous weapons use and certain surveillance activities involving Americans, although the companies and government agencies have not fully disclosed how those limits apply in different military scenarios.
Comments
0No comments yet. Be the first to comment.