A security breach at the maker of ChatGPT last year revealed internal discussions among researchers and other employees, but not the code behind OpenAI’s systems.