OpenAI Employs Human Contractors to Review ChatGPT Conversations, Raising Privacy Concerns
OpenAI is hiring hundreds of contractors to review real user conversations with ChatGPT, a practice known as 'Project Lily,' according to 404 Media. These contractors analyze prompts and entire chat histories, which can include sensitive personal information. The primary goal of these review teams is to improve ChatGPT's responses, with contractors rating and critiquing the chatbot's generated replies. Internal documents indicate that reviewers are training ChatGPT to avoid anthropomorphizing itself and to be less sycophantic, issues that have been linked to past incidents, including lawsuits alleging that an overly sycophantic AI model contributed to suicides. While OpenAI states that contractors do not see usernames and efforts are made to remove personal information, the company acknowledges that sensitive details can still reach reviewers. This practice challenges the common user perception of private, one-on-one interactions with the chatbot.
