404 Media reported on Tuesday that OpenAI removed contractors who leaned on chatbots for their own work. The job is to read real ChatGPT conversations and grade the answers by hand. Graders were trained to spot fakes by repeated wording, too many dashes and suspiciously fast work.