OpenAI Reportedly Pays Contractors $50+ an Hour to Review ChatGPT Chats

Leaked internal materials reveal that OpenAI contractors review some ChatGPT conversations as part of the company's model improvement efforts. Under a project called "Project Lily," reviewers grade responses, summarize user intent, and flag issues like robotic phrasing. OpenAI uses automated privacy filtering to remove identifying details before conversations reach reviewers, but the system may not catch every rare identifier or ambiguous reference.

Contractors earn over $50 an hour to perform this work through intermediary firms. While OpenAI acknowledges the privacy filter can fail on rare cases, contractors are not aware that their work may involve reviewing user conversations. The disclosure may lead to increased privacy and regulatory scrutiny over how consumer AI conversations are processed and disclosed.