Work, learning and AI · 6 min read

ChatGPT privacy settings

By default, ChatGPT may use conversations from personal accounts to improve its models, but this can be turned off with a single toggle. For a single sensitive conversation there is temporary chat, and for links already sent there is the list of shared links.

In this article
  1. Opting conversations out of model training
  2. Temporary chat: what is it for?
  3. Memory: what does ChatGPT remember?
  4. Managing shared conversation links
  5. Downloading and deleting data
  6. Privacy rights in the EU
  7. What else to watch for?
  8. If it doesn't work

Opting conversations out of model training

For Free, Go, Plus and Pro accounts, conversations, uploaded files and images may be used to improve the models, depending on the setting. This is how to turn it off:

  1. On the web, open the account menu (profile icon) and choose “Settings”.
  2. Open “Data controls”.
  3. Turn off the “Improve the model for everyone” toggle, then click “Done”.
  4. In the iOS and Android apps, open the sidebar and tap the profile icon to reach the settings; model training can be turned off in the same place, under “Data controls”.
  • For a logged-in account, the setting applies on every device. When using ChatGPT logged out, it has to be set in each browser.
  • Turning it off does not delete or hide earlier conversations: history, archived chats and projects remain.
  • The same request can also be made on the OpenAI Privacy Portal (privacy.openai.com), with the “Do not train on my content” option. Either of the two methods is enough.
  • OpenAI does not use content from work and education plans (Business, Enterprise, Edu) for model training by default.
Toggle or temporary chat?
Should only one conversation be opted out, not all of them?
Yes, just one
  1. Start a new chat in temporary mode
  2. It does not appear in the history and does not update memory
  3. For safety reasons, it may be kept for up to 30 days
No, all of them
  1. “Settings” → “Data controls”
  2. Turn off “Improve the model for everyone”
  3. The history is kept

Temporary chat: what is it for?

Temporary chat is a kind of incognito mode: it has to be turned on when starting a new conversation, before the first message is sent.

  • It does not appear in the chat history or in search, and it cannot be reopened after leaving it.
  • It does not create or update memories.
  • As long as it stays temporary, it is not used for model training. For safety reasons, OpenAI may keep a copy for up to 30 days.
  • When starting it, there is a choice of whether it should take existing memory and custom instructions into account (personalized or unpersonalized mode). This cannot be changed once the conversation has started.
  • If a temporary conversation needs to be kept, it can be converted into a normal conversation; from then on, the account's usual settings apply to it.
  • If a GPT sends data to an outside service during the conversation, that service's data practices apply, and the data may be kept there for longer than 30 days.

Memory: what does ChatGPT remember?

Memory is used to personalize answers and is independent of the model training toggle. Where: “Settings” → “Personalization” → “Memory”.

  • “Reference saved memories”: using facts ChatGPT has remembered (for example a name or preferences). Turning it off does not delete existing conversations.
  • “Reference chat history”: referring to earlier conversations in answers.
  • “Memory summary”: shows what ChatGPT knows about the user; individual items can be deleted from here. ChatGPT can also be asked directly what it remembers.
  • Deleting a conversation does not necessarily delete a saved memory created from it. Both the memory and the source conversation have to be deleted.
  • The “Delete and turn off memory” option deletes the items shown in the summary and turns memory off in one step; it does not delete conversations.

A conversation can be sent as a link with the share option. It is important to know how this works:

  • For personal accounts, the link is a snapshot: messages written after sharing do not appear in it until the link is updated.
  • Anyone who receives the link can open it and pass it on. Search engines are not supposed to index it, but that does not make the content private.
  • The link does not expire on its own. A shared conversation is anonymous by default, but personal data mentioned in the text of the conversation remains visible.
  • Deleting one link: “Settings” → “Data controls” → “Manage” next to “Shared links”, then delete the link in the list.
  • Deleting all links: in the same place, with the option to delete all shared links.
  • If the original conversation is deleted, its shared link stops working as well.

Downloading and deleting data

  1. Requesting a data export: “Settings” → “Data controls” → “Export”, then “Confirm export”.
  2. An email or text message arrives when the export is ready (this can take up to 7 days). Click “Download data export” in the message; the link is valid for 24 hours and only works while logged in to the account that made the request.
  3. The export is a ZIP file with the conversations and account data. Deleted conversations are not included.
  • Conversations can be deleted one by one or all at once. OpenAI removes deleted conversations from its systems within 30 days, unless they have to be kept for legal or security reasons.
  • Deleting the account is permanent and also affects its data and subscriptions. A subscription started in the App Store or on Google Play has to be canceled separately first: Managing and canceling a ChatGPT subscription.
  • For work accounts (Business, Enterprise), export and account deletion are handled through the workspace owner.

Privacy rights in the EU

  • For users living in the European Economic Area, the data controller is OpenAI Ireland Limited (Dublin). This is stated in OpenAI's European privacy policy.
  • Access to data, correction, deletion, portability and restriction can be requested, and objections can be raised to processing based on legitimate interest.
  • Requests can be submitted on the privacy.openai.com portal or at [email protected].
  • Complaints can be made to the Irish Data Protection Commission or to the data protection authority of the user's own country.

What else to watch for?

  • Passwords, bank card numbers, health data or other people's personal data are best not entered, even with model training turned off.
  • Memory and model training are two separate settings: turning one off does not turn off the other.
  • On a shared computer, log out after use. About the login methods: Logging in to ChatGPT on the web and in the app.

If it doesn't work

If I turn off model training, will my earlier conversations disappear?

No. The toggle only controls whether conversations can be used to improve the models; history and projects remain.

Is a temporary chat really seen by nobody?

It does not appear in the history and is not used for model training, but for safety reasons OpenAI may keep a copy for up to 30 days.

Can a shared link that has already been sent be revoked?

Yes: the link can be deleted with the “Manage” button next to “Shared links”, after which it can no longer be opened. However, nothing can be done if someone saved or took a screenshot of the content earlier.

Does deleting a conversation also delete the information from memory?

Not necessarily. The saved memory has to be deleted separately under “Personalization” → “Memory”.

How long does it take for deleted conversations to be removed?

OpenAI removes deleted conversations from its systems within 30 days, unless they have to be kept for legal or security reasons.

© 2026 login-start.com · Unofficial guides to popular services