OpenAI hires contractors to review ChatGPT chats under project Lily
Turns out, OpenAI has been hiring contract workers to check out real ChatGPT conversations as part of something called "Project Lily."
These reviewers look at how clear and helpful the AI's replies are, and whether they follow the rules.
But here's the catch: some of these chats include personal or sensitive information, raising questions about privacy.
ChatGPT privacy filter may fail
OpenAI uses a "Privacy Filter" to hide sensitive details in your chats, but it doesn't always work.
The company didn't make this super clear at first, though they've since pointed to a section on its website that says humans may review some content.
If you'd rather not have your chats used to improve the model for everyone, you can turn off the "Improve the model for everyone" setting in ChatGPT, just a quick toggle in your settings.