OpenAI blocked 15,000-user effort to copy hidden model reasoning
OpenAI says it disrupted a campaign that exposed protected model reasoning at scale, a warning for anyone building or operating AI systems that model outputs can be weaponized for unauthorized copying. Operators manipulated conversations rather than breaking encryption or accessing stored chats, including copying encrypted reasoning between conversations and asking another model to decrypt it. OpenAI saw 16,000 extraction requests from more than 4,000 users on July 24 and 25, then found related activity across more than 15,000 users and disrupted it by July 28.