404 Media Exposes OpenAI's 'Project Lily': Humans Reading Your ChatGPT Chats

On September 14, 404 Media published an investigation, 'Inside Project Lily: The Humans Reading Your ChatGPT Chats,' based on leaked internal documents and real user prompts, revealing that OpenAI contractors read user-ChatGPT conversations — meaning such chats are not fully private. The report spread quickly across the AI community, with no direct OpenAI response as of posting.

Confirmed

• According to 404 Media, OpenAI's 'Project Lily' employs hundreds of contractors to read large volumes of real user-ChatGPT conversations, including prompts, to rate and improve model responses.
• These conversations may contain highly private personal information.
• The report details the reviewers' working conditions and psychological impact of exposure to intimate conversations.
• Consumer plans default to sharing chats for training, covering most of over 900 million users.

Why it matters

• Users generally assume AI chats are private; the report directly breaks that expectation, touching core disputes over privacy boundaries and data use.
• Reposters including RebeccaBellan, VoidStateKate and lilyraynyc voiced concerns — VoidStateKate said she had long expected this. The impact on user trust warrants continued attention.

2026-09-14 ~ 2026-09-16 · 7 related posts

• [source] 404 Media: Inside 'Project Lily,' the Humans Reading Your ChatGPT Chats — 404mediaco · 2026-09-14
• [source] Inside Project Lily: OpenAI Hires Humans to Read Your ChatGPT Chats — 404mediaco · 2026-09-14
• OpenAI hired contractors to read real ChatGPT chats, per 404Media — VoidStateKate · 2026-09-15

4 near-duplicate retellings: esporx · RebeccaBellan · lilyraynyc · RareEarthPetal