18/09/2026
A stranger may already have read a ChatGPT conversation you thought was private.
Not OpenAI staff in a glass office. Hundreds of outside contractors, paid to open real conversations, summarise what was asked, and grade the replies. 404 Media reported it this week under an internal programme called Project Lily.
Usernames are stripped and a privacy filter runs first. OpenAI says that filter can miss things.
Now look at the screenshot. That's my ChatGPT. I can't switch that setting on even if I wanted to, because our workspace at Medialogix has made the decision so nobody on the team has to remember to.
On a personal account, Free, Plus or Pro, it's on by default. You turn it off yourself.
Same logo on the screen. Completely different deal on your data.
So we don't ask people to be careful. Being careful is a hope, not a system.
Twenty seconds on your phone: Settings, then Data controls, then Improve model for everyone. Switch it off. It only stops new chats, not the ones already in there.
Banning AI isn't the answer. I want my team using more of it, not less. Safely.
Has your business actually got an AI policy? Or is everyone making up their own rules?
We wrote one for our team at Medialogix. It's yours if you want it.
Comment POLICY and I'll send it over.