I’ll be honest, this one made me stop and scroll back through my own ChatGPT history. Health questions, money worries, half-finished work drafts, things I’d never type into a search bar. Most of us treat ChatGPT like a private journal, and it isn’t one by default.
I came across a really useful post from a LinkedIn creator who put it bluntly: “You’ve told ChatGPT things you’d never tell Google.” The author’s point is simple. Unless you change the settings yourself, your chats can be used to train the next model. So the post lays out a step-by-step cleanup to erase your data from OpenAI’s servers and lock things down from here on.
The part I liked most is the legal angle. As the original poster points out, under GDPR, OpenAI is legally required to delete your data when you ask. You don’t need any special access or a lawyer. You just need to know where the buttons are.
🧹 Why this cleanup matters
Before we get to the steps, here’s some quick context. ChatGPT stores several kinds of data about you, and each lives in a different place. Your chat history, shared links, saved memories, training permissions and file storage are all separate. One toggle won’t cover all of them.
That’s why a full checklist helps so much. If you skip one area, some of your data stays put. The expert’s list covers each spot one by one, and I’ll add why each step matters as we go.
🔐 The step-by-step cleanup
Work through these in order. The first few clean up what’s already there. The rest stop new data from piling up.
- Delete the individual chats. Click the 3 dots next to any chat in the left sidebar and hit “Delete.” The author admits this is obvious and slow, but it works: the chat is permanently deleted from OpenAI’s systems within 30 days.
- Delete the chat links you’ve shared. Go to ChatGPT Settings, then “Data Controls,” and delete every shared chat on the list. Deleting these links does not delete your own chats.
- Ask ChatGPT what it knows about you, then draft a deletion request. The creator uses a set of 3 prompts, run in order in the same chat. The first shows what ChatGPT keeps on you, the second walks you through every setting, and the third writes a formal deletion request.
- Turn off data training. Go to Settings, then Data Controls, and switch off “Improve the model for everyone.” It’s on by default, which means your chats are used to train the model.
- Turn off Memory. Go to Settings, then Personalisation, and manage your ChatGPT memory there. According to the post, memory is used for training unless the toggle is off.
- Turn off Codex memory. If you use Codex, it has its own training switch in Codex Settings. The main ChatGPT toggle doesn’t cover it.
- Use Temporary Chats. The post calls this the number 1 rule. Use Temporary Chat (upper right corner) for anything personal. OpenAI still keeps that data for 30 days for safety, but it’s never used for training.
- Clean up Storage. Go to Settings, then Storage, and choose to clean up old chats and review them.
- Go through every setting right away. The author’s last nudge is to do this now, not “someday.”
Each step closes a different door. Deleting chats handles the past, the toggles protect the future, and Temporary Chats give you a safe space for sensitive questions.
💡 Why each step matters
Some of these steps are easy to underrate, so here’s my quick take on the important ones.
- Shared links: These are the sneaky ones. Anyone with the link can read that conversation, even if you forgot you shared it months ago.
- The training toggle: This is the biggest single switch. Turning it off stops your future conversations from feeding the next model.
- Memory: ChatGPT’s memory builds a profile of you over time, like your job, family, preferences and projects. Reviewing it is eye-opening. You might be surprised at what it saved.
- Codex: Developers often paste code, API details and company logic into Codex. Since it has a separate switch, it’s easy to miss.
- Temporary Chats: Think of this as incognito mode for ChatGPT. It’s a habit, not a one-time fix.
📝 Prompt of the Day
The original prompts live in the full post, but you can try the core idea right now. Open a new chat and ask something like this:
“Based on everything you remember about me, list every personal detail you have saved. Group it by category and tell me where in Settings I can delete each item.”
Then follow up with: “Write a short, formal data deletion request under GDPR asking OpenAI to erase all personal data linked to my account.”
You can submit that request through OpenAI’s privacy request portal. Just make sure you’re logged in with the same account you use for ChatGPT.
🛡️ Tips to stay private going forward
Cleaning up once is great. Keeping things clean is even better. Here’s how I’d turn the author’s checklist into a habit:
- Set a monthly reminder to review Memory and Storage. It takes 5 minutes.
- Default to Temporary Chat for health, money, legal or relationship questions.
- Strip identifying details before pasting documents. Swap names, addresses and account numbers for placeholders.
- Check settings after big updates. New features like Codex sometimes ship with their own separate toggles.
- Repeat on every device and account. If you have a work account and a personal one, each needs its own cleanup.
This matters beyond ChatGPT too. AI assistants are becoming part of how we think, plan and work. That makes privacy settings just as important as the passwords we already guard. I think posts like this one are valuable because they turn a vague worry into a 15-minute task anyone can finish.
Huge thanks to the person who shared it for putting this together so clearly. Check out the full LinkedIn post for the complete walkthrough and the original 3 prompts, then go flip those toggles today!