Two California ChatGPT users sued OpenAI in federal court in a proposed class action, alleging the company never clearly told people that outside contractors in a program called Project Lily read their conversations to help improve its models. The complaint says hundreds of contractors can see real prompts, sometimes whole chats, then summarize what the user wanted and score ChatGPT’s replies, and that an automated filter misses some personal details in chats about health, money and legal problems. OpenAI does tell users its own staff can check flagged content and that business admins can see account data, but the plaintiffs argue its privacy policy names hosting, payment and support vendors without mentioning annotation or evaluation contractors.
