404media.co via Reddit

404 Media: OpenAI 'Project Lily' Hires Hundreds of Contractors to Read Real ChatGPT Prompts

By Alexis Dufresne Published September 14, 2026 at 16:58 UTC

OpenAI AI Ethics Surveillance ai-business

Summary

404 Media details 'Project Lily,' OpenAI's internal codename for hiring hundreds of contractors to read real ChatGPT users' prompts—many containing sensitive personal information—and rate the responses on a 1-7 scale. Contractors are recruited by Crossing Hurdles and paid via Mercor (one worker reports over $50/hour). ChatGPT has more than 900M users and model training is turned on by default for consumer plans; the piece notes Anthropic and Google run similar human-review programs.

Shared on Bluesky by 10 AI experts (top 5 by trust)

404 Media @404media.co: NEW: EXCLUSIVE: Humans are reading ChatGPT users’ prompts to improve OpenAI’s models, and those chats can include sensitive, personal inform… →
Joseph Cox @josephcox.bsky.social: Your ChatGPT chats are not private, and humans are reading them. Anthropic's Claude too. More here: www.404media.co/inside-proje... →
Sarah T Roberts, PhD @ubiquity75.xyz: It’s always humans, all the way down. This is the social media playbook, on steroids. →
Luke Stark @lukestark.bsky.social amplified

Joseph Cox @josephcox.bsky.social

There is a lot of new info in this article, based on documents, Slack chats, instruction guides, and me seeing real ChatGPT user prompts in the review system. Firstly, a worker we spoke to thinks ChatGPT users have no id…
View on Bluesky →
Jason Koebler @jasonkoebler.bsky.social: OpenAI is paying real people to read live ChatGPT prompts and responses, often containing very sensitive information, in part to determine i… →

Originally reported by 404media.co

Read the original article →

Original headline: 404 Media: OpenAI 'Project Lily' Hires Hundreds of Contractors to Read Real ChatGPT Prompts