OpenAI pays contractors $50+/hour to read full ChatGPT conversations, 404 Media reveals
thisguyknowsai · x · 2026-09-26
404 Media's "Project Lily" report reveals OpenAI pays hundreds of outside contractors over $50/hour to review real ChatGPT conversations — not just messages but whole chats, sometimes with a "user memories summary" that can hint at where users live.
Key points:
- Model training is on by default for Free, Plus, and Pro; opting out only applies to new chats
- Chats pass a privacy filter first, but OpenAI admitted it can miss personal details in shorter conversations
- Some users explicitly asked ChatGPT to keep secrets; OpenAI didn't directly answer where users are warned humans may read chats
- A class action was filed in San Francisco two days after the report
- Anthropic also confirmed it uses human review
The author's take: human review isn't evil, but 900 million people assume they're talking to a machine and the settings page doesn't say otherwise.
More from Companies & People
- Data Suggests Median OpenAI Employee Uses AI About as Much as Average Twitter User — xeophon · 2026-09-26
- Free Deep Learning Course by an Oxford/DeepMind Professor Circulates Amid Claims of $650K Salaries — NandoDF · 2026-09-26
- Musk: SpaceX and Tesla training data are Grok's edge in real-world engineering — XFreeze · 2026-09-26
- Google DeepMind CTO Koray Kavukcuoglu still has no Wikipedia page — HaydnBelfield · 2026-09-26
- Former OpenAI researcher: well-executed Google Muse could outgrow search — infoxiao · 2026-09-26
- Musk calls Grok 4.7 a solid workhorse, expects xAI to catch up to frontier next year — XFreeze · 2026-09-26