ChatGPT privacy settings: how to stop it training on your chats

For sellers, freelancers and anyone who pastes customer messages, supplier prices or business ideas into ChatGPT.

By the end you will have switched off model training on your account, you will know the three cases where your text can still be used, and you will have a short rule for what never goes into a chat.

Everything below comes from OpenAI's own help pages on data controls and Temporary Chat.

Why this matters

On personal Free, Plus and Pro accounts, sharing your conversations to improve the model is on by default. If you paste a customer's name, phone number and address to get a reply drafted, that text is part of what can be used. One setting changes this for new conversations.

The setting, step by step

1. Open ChatGPT on the web and click your account menu.
2. Choose Settings.
3. Choose Data controls.
4. Switch off "Improve the model for everyone".
5. Select Done.

On the iPhone or Android app: open the sidebar, tap your profile icon, tap Settings, then Data controls, and switch off the same option.

If you are signed in, the setting follows your account across devices. If you use ChatGPT signed out, the setting is saved only in that browser.

OpenAI also offers a second route through its Privacy Portal, an option called "Do not train on my content". Either one is enough.

What the setting does

• New conversations are not used to train OpenAI models.
• Your chats still appear in your history. You keep your history and lose nothing.

The three cases it does not cover

1. Feedback. On individual plans, if you press thumbs up or thumbs down on a response, OpenAI may use the whole conversation attached to that feedback, even if you opted out. If the chat contains private details, skip the thumbs.
2. Older chats. The switch applies to new conversations. Delete what you do not want to keep.
3. Signed-out use. If you use ChatGPT without logging in, the setting is saved only in that browser, so log in on every device you use.

Temporary Chat for sensitive work

Temporary Chat is a separate mode for one-off, sensitive questions.

• OpenAI says a temporary chat is not used to improve its models while it stays temporary.
• It does not appear in your history and does not create or update memories.
• OpenAI may keep a copy for up to 30 days for safety reasons.
• If you save a temporary chat, it becomes a normal chat and your model-improvement setting applies from then on.

A paste rule for sellers

Switching the setting off protects future chats. The safest data is the data you never paste. Use this rule before you hit send:

• Replace customer names with Customer 1, Customer 2.
• Remove phone numbers, addresses and order IDs. Write "phone" or "address" instead.
• Keep supplier prices and margins out unless you need them for the answer.
• Use Temporary Chat when the question is about something you would not post publicly.

Copy-paste prompt for customer replies without personal details:

I run a COD store in Morocco. I will describe a customer situation without names or numbers. Write a short, polite WhatsApp reply in Darija (Latin letters). Situation: [describe what the customer said and what you want to happen]

Replace the text in brackets with your own situation.

Quick audit

Do this once on every device where you use ChatGPT:

1. Settings, Data controls: "Improve the model for everyone" is off.
2. You know where Temporary Chat is.
3. You do not press thumbs on chats with private details.
4. Old chats with customer data are deleted.

My honest take

This takes a minute and it removes a default you probably never chose. It does not make ChatGPT private, and it does not undo what you already pasted. Treat it as a seatbelt: set it once, then keep the sensitive details out of the chat anyway.