Turn Off ChatGPT Training and Improve Your Privacy

You can stop new ChatGPT conversations from being used to improve OpenAI’s models by changing the data control setting in your account. This guide shows where to find it, how to use Temporary Chat, and what else to check to reduce the personal information you share. The exact labels may vary slightly by app or account type.

What You’ll Need

Access to your ChatGPT account in a web browser or the official mobile app.
A moment to review the setting on each account you use. Personal, work, and school accounts may have different controls.

Step-by-Step

Open ChatGPT and sign in. Use the account whose conversations you want to control. On the web, open ChatGPT and select your profile or account menu. In the mobile app, open the side menu or tap your profile icon, depending on the current layout.
Go to Settings. Choose Settings from the account menu. If you do not see the menu, look for the profile icon, initials, or a menu button. Update the app if its layout is substantially different from the instructions.
Open Data Controls. In Settings, select Data Controls. Find the control labeled Improve the model for everyone or a similar model-improvement option. OpenAI may change wording or organize controls differently over time.
Turn off model improvement. Switch the control off and confirm if prompted. Check that it remains off after closing and reopening Settings. This is the key step for opting out of using your new chats to improve models; it does not erase chat history.
Repeat on other accounts or devices if needed. The preference is associated with the account, but check any separate personal, work, or school accounts you use. If a setting appears to differ between the web and app, verify the signed-in email and account before assuming it has synced.
Use Temporary Chat for conversations you do not want in your history. Start a new chat and choose Temporary Chat, typically from the model or chat controls. Temporary Chats do not appear in your regular history and are not used to improve models. OpenAI may retain them for a limited period for safety or abuse monitoring, so do not treat this mode as a promise of immediate deletion or absolute confidentiality.
Review existing conversations separately. Turning off model improvement is not the same as deleting chats. To remove an individual conversation, use its menu in the chat list and select the delete option. To clear or manage more history, review the available chat-history controls in Settings. Read the confirmation carefully; deletion and data retention can be subject to the service’s stated policies.
Reduce sensitive information in future prompts. Avoid entering passwords, payment details, private keys, confidential workplace material, or personal information you do not need to share. Remove names and identifying details when possible. Check any connected apps, shared links, or custom instructions that may contain information you no longer want associated with your account.
Check account-specific controls. If you use ChatGPT through an organization, your administrator may manage data settings, and the consumer-account steps may not apply. Ask the workspace administrator which retention and model-improvement rules apply. Do not move confidential work data into a personal account to get around workplace controls.

Common Mistakes

Assuming turning off training deletes old chats. It controls model improvement for new conversations; use the chat’s delete control to remove history you no longer want visible.
Assuming Temporary Chat means no retention. It keeps the conversation out of regular history and excludes it from model improvement, but limited retention may still apply for safety purposes.
Changing the wrong account. People often have separate logins or switch between personal and organizational workspaces. Check the account menu before changing settings.
Confusing chat history with training. A conversation can remain in your history even when model improvement is disabled. Conversely, deleting a chat should not be treated as a substitute for changing the data control.
Sharing sensitive information because a setting is off. Opting out is not a guarantee that a service is suitable for confidential data. Follow your employer’s rules and the service’s privacy information.

Tips

Recheck Data Controls after major app updates or when switching accounts, especially if the setting is important to your workflow.
Use Temporary Chat selectively for one-off conversations that you do not need in your normal history, while remembering its retention limits.
Before submitting a prompt, replace real names, addresses, account numbers, and other identifiers with placeholders where practical.
Review shared conversation links and connected services. A link you created or an integration you enabled may expose information separately from chat history settings.
For work or school accounts, consult the organization’s administrator or approved guidance rather than relying on consumer settings.

Troubleshooting

The model-improvement switch is missing: Confirm you are in the correct account and open Settings then Data Controls. Try the web version or update the app. If the account is managed by an organization, some controls may be unavailable or administered centrally.

The setting turns back on or does not appear to save: Refresh the page, sign out and back in, then check the control again. Make sure you did not switch workspaces or accounts. If the issue persists, consult OpenAI’s current Help Center or contact support through the service.

Temporary Chat is unavailable: Look in the new-chat controls or model selector and check for an app update. Feature availability and placement can vary. If it is not offered on your account, avoid entering sensitive details rather than assuming a normal chat is temporary.

You cannot delete a conversation: Check that you are viewing the right account and use the conversation’s menu in the history list. If the delete option fails, reload or try another supported client. For account-wide data requests, use the privacy or data controls provided by OpenAI and read the confirmation before submitting.

Frequently Asked Questions

Does turning off model improvement delete my existing chats?
No. The setting controls whether new conversations are used to improve models; delete individual chats separately if you want them removed from your visible history.

Will my chats still appear in the sidebar after I opt out?
Usually, yes. Opting out of model improvement does not by itself disable or clear chat history.

Are Temporary Chats saved?
They do not appear in regular history and are not used to improve models, but they may be retained for a limited period for safety or abuse monitoring.

Does the setting apply to every ChatGPT account I use?
Do not assume so. Check each separate account or workspace, especially if you use both personal and organization-managed access.

Can I use ChatGPT privately by turning off training?
The setting reduces one use of your conversations, but it is not a guarantee of complete confidentiality or zero retention. Avoid sharing information that should remain private.

Can I change my mind later?
You can return to Data Controls and change the available preference. Check the current setting after you change it; a later change should not be assumed to alter how earlier conversations were handled.

General steps for current mainstream versions; menus can differ by device and version. See the disclaimer or report an error.

More in this bed

The whole Tech How-To bed →