[OpenAI Recruits Hundreds of Outsourced Reviewers to Manually Examine ChatGPT User Conversations for Model Optimization] According to 404 Media, OpenAI is hiring hundreds of outsourced reviewers to continuously read and label real user conversations with ChatGPT, aiming to improve the quality of model responses. These review tasks include complete conversations between users and ChatGPT, some of which contain personal privacy and other sensitive information. Internal documents reveal that reviewers are being used to train ChatGPT to reduce anthropomorphism and minimize "flattering" responses. Previously, OpenAI's GPT-4 model was accused in multiple lawsuits of excessive pandering, allegedly linked to user suicides. OpenAI stated that it would attempt to remove personal information before submitting conversations for review but acknowledged that sensitive content might still be retained. Anthropic has also confirmed the use of similar manual review mechanisms to train its models.