OpenAI Lets Contractors Read ChatGPT Prompts by Default — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI internal documents detail contractors evaluating anonymized ChatGPT prompts and chats to improve its models, and 404 Media reports those reviewed conversations can include sensitive user content
- ChatGPT consumer plans have model training toggled on by default, meaning user prompts feed into OpenAI's model-improvement process unless users actively opt out
Why it matters: The leaked docs confirm that consumer ChatGPT model training is opt-out, not opt-in, and that the anonymized chats contractors review can include sensitive material. Users who never navigated to the data-controls setting are having their conversations read by humans as part of OpenAI's training pipeline.
Ask SkimNews
