The Pentagon reportedly asked OpenAI to provide an AI that doesn't like saying no to requests
Documents revealed disputed Pentagon requests for an OpenAI system with fewer refusals, while officials denied such language existed in final agreements.
Recently released FOIA documents have sparked a dispute over Pentagon plans for an AI system with less restrictive guidelines. OpenAI has vehemently denied agreeing to a contract that would limit the AI's refusal rates during military applications. The controversy emerged after a Pentagon lawyer claimed the contract included language requiring minimal refusal rates from models, but later retracted that statement.
The documents obtained through a Freedom of Information Act lawsuit described OpenAI's models intended for national security work that would require minimal refusal rates. However, officials from both organizations maintain that the disputed language never made it into the final agreement. OpenAI spokesperson Nate Evans stated that the document in question was an earlier draft, not the executed contract.
The Pentagon, on the other hand, maintains that the phrase "minimal refusal rates" does not appear in any active contract. The disagreement has reignited discussions about the level of control technology companies should retain over AI used by governments, especially in defense applications.
Written by urgent.news from TechRadar's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
