Now, OpenAI CEO Sam Altman says the company has fixed the problem. In a post on X, Altman writes, “If you tell ChatGPT not to use em-dashes in your custom instructions, it finally does what it’s ...