The U.S. Department of Defense Allegedly Seeks AI with the 'Lowest Rejection Rate' for Military Use

Huo Xing Cai Jing reports that documents obtained by The Intercept through a Freedom of Information Act lawsuit reveal that the U.S. Department of Defense requested OpenAI to develop an artificial intelligence version tailored for military use with an “extremely low refusal rate” to commands. This phrasing appears in Contract P00003, an updated agreement for a two-year prototype project worth up to $200 million, covering logistics, intelligence decision-making, and operational applications. OpenAI, Google, xAI, and Anthropic all agreed in 2025 to develop such militarized prototypes. OpenAI spokesperson Nate Evans stated that the company never agreed to contract terms containing “extremely low refusal rate,” asserting that the document in question was an early draft and that the final executed contract contains no such language. Department of Defense spokesperson Jacob Bliss also confirmed that the phrase does not appear in any current contract. The Intercept noted that the obtained documents were not labeled as drafts; a Justice Department attorney representing the Pentagon initially confirmed the document as an executed version, then later retracted that confirmation and requested further investigation. The most recent contract, signed on February 27, permits OpenAI services to be deployed on military classified networks, with identical delivery clauses fully redacted. OpenAI claims it has established clear boundaries against autonomous killing and surveillance targeting U.S. citizens, yet the contract text permits any use deemed lawful by the government.