ChatGPT Privacy Settings: History, Training and Memory

Yes—ChatGPT lets you turn off model training, while chat history and memory are separate controls that work differently. To keep conversations out of training, switch off “Improve the model for everyone”; to limit what ChatGPT remembers, manage or disable Memory, and use Temporary Chat when you want a conversation that does not appear in your history.

What each privacy setting controls

“Chat history,” “model training,” and “Memory” are related privacy topics, but they are not one master switch. Turning off one does not automatically turn off the others.

Model training: The “Improve the model for everyone” control governs whether eligible conversations may be used to improve OpenAI’s models. Turning it off does not remove conversations from your chat list.
Chat history: Your normal conversations are saved in your account history unless you delete them or use a chat mode designed not to save them to that history. History is useful for finding and continuing earlier conversations.
Memory: When enabled, ChatGPT can retain selected details, such as your preferred writing style, to personalize future chats. Memory is distinct from the transcript of a particular conversation.

These controls may appear under different names or locations as ChatGPT’s interface changes, and availability can vary by account, plan, region, or workspace. Check the current settings shown in your own account rather than assuming every account has identical options.

How to turn off model training

In ChatGPT, open your profile menu and go to Settings, then look under Data Controls for Improve the model for everyone. Turn it off. The wording and exact placement may change, but the setting is intended to apply to your account across devices when you are signed in.

With this setting off, conversations can remain visible in your history and available to you, but they are not used to improve models under that setting. This is not the same as deleting conversations, and it does not mean the conversation is never processed to provide the service. OpenAI may retain and process information as described in its privacy policy and applicable product terms.

There can also be separate choices for feedback. If you submit a rating or other feedback on a response, the associated conversation may be handled differently for model improvement. Read the notice shown when providing feedback, especially if you have disabled training and want to understand what sharing a rating entails.

How to manage Memory

Look in Settings for Personalization and Memory. Depending on what is available for your account, you may be able to turn memory on or off, review saved memories, or remove individual items. You can also ask ChatGPT what it remembers or tell it to forget a detail, then verify the result in Memory settings.

Switching Memory off is not necessarily the same as erasing memories already saved. If you want to remove information, review and delete the saved memories as well. Also consider deleting the chat where you originally shared the detail: saved memories and chat transcripts are managed separately, and deleting one does not necessarily delete the other.

For example, you might delete a conversation in which you mentioned a dietary preference, but the preference may still be present as a saved memory. Conversely, deleting a saved memory does not necessarily remove the original message from your chat history. Check both places when you want a detail removed from both.

When to use Temporary Chat

Temporary Chat is designed for a conversation that should not be added to your regular chat history or create and use personalization memories. It is useful for a one-off question you do not need to revisit—for example, asking for help rewriting a generic paragraph without saving that exchange to your chat list.

Temporary does not mean zero data handling or guaranteed immediate erasure from every system. OpenAI says temporary chats may be retained for a limited period for safety purposes. The exact treatment is governed by current product explanations and privacy terms, so do not use Temporary Chat as a promise that no record can exist.

Temporary Chat also does not make it wise to share sensitive information. Avoid entering passwords, payment details, private identifiers, confidential work material, or information about someone else unless you have a clear reason and permission to share it.

Choose the right combination

Pick settings based on what you want to achieve. Common combinations include:

Keep chats, opt out of model improvement: Turn off “Improve the model for everyone.” Your normal chat history can still be available.
Keep history but reduce personalization: Turn off Memory and remove any saved items you do not want retained. This does not itself disable model improvement.
Do not keep a conversation in normal history: Start a Temporary Chat. It is separate from the training control, so review the current explanation of how temporary chats are handled.
Remove an old conversation: Delete it from chat history. If it contributed a saved memory, remove that memory separately.

For work or school accounts, an administrator may manage workspace settings or impose policies. Personal account controls may not be available in the same way, and organizational data handling can differ. Check with the workspace administrator if a control is missing or locked.

Common misconceptions and limits

A frequent misconception is that turning off chat history also disables model training. They are different controls: the training opt-out does not automatically erase or hide existing conversations. Another is that deleting a chat automatically clears Memory; saved personalization details may need separate removal.

Privacy controls also do not turn ChatGPT into an encrypted private notebook or guarantee that no human or automated safety process can ever review information. They govern specific product uses and storage behaviors, not every possible form of processing. For confidential or regulated information, follow your organization’s rules and the service terms that apply to your account.

If you cannot find a setting, confirm that you are signed into the intended account, open the current Settings menu, and check whether you are using a managed workspace. Interface changes or account availability can explain differences from older instructions. Avoid relying on screenshots or menu paths that may be out of date.

Frequently Asked Questions

Does turning off ChatGPT chat history stop model training?
Not necessarily. History and model improvement are separate controls. To opt out of model improvement, turn off “Improve the model for everyone” in Data Controls.

Can I turn off ChatGPT memory but keep my chats?
Yes. Memory and chat history are separate. You can disable Memory while keeping normal conversations in your history, subject to the options available in your account.

Does deleting a chat delete a saved memory?
Not always. A detail saved as Memory can remain after you delete its original conversation. Review both chat history and Memory settings to remove both copies.

Are Temporary Chats completely private?
No setting should be read as a guarantee of zero processing or immediate erasure. Temporary Chats are not added to regular history and do not create or use personalization memories, but may be retained for a limited period for safety purposes.

Does opting out of training delete my old chats?
No. It changes whether eligible conversations are used for model improvement under that control; it does not delete your history. Delete conversations separately if you want them removed from your chat list.

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 →