OpenAI employs contract workers to review ChatGPT conversations for model improvement. Workers rate chats on a 1‑7 scale to reduce flattery and overly human‑like responses. Anonymized prompts may still contain sensitive data; users can opt out by disabling the “Improve the model for everyone” setting, which is on by default.
Opening Kapyn…