Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →To stop new conversations on a personal ChatGPT account from being used to improve OpenAI’s models, open the account menu, choose Settings, open Data controls, switch Improve the model for everyone off, and select Done. The setting applies across devices where you are signed in. It does not delete existing chats or remove them from your history.
For a conversation that should not appear in history or create or update memories, start a new Temporary chat before sending your first message. Temporary chats are not used to improve models while they remain temporary, although OpenAI may retain a copy for up to 30 days for safety.
As an Amazon Associate I earn from qualifying purchases.
What the ChatGPT opt-out actually controls
“Data collection” can mean several different things. The control documented by OpenAI is specifically the model-improvement setting. Turning it off tells OpenAI not to use your new conversations to improve its models. It is not a universal switch that stops every form of processing required to operate an account, provide a reply, protect the service, or apply a workspace policy.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11With the model-improvement setting off, conversations you start afterward can still appear in your chat history. Existing chats remain visible, including chats you have archived or placed in a project. If you need a conversation removed, delete that chat or the account separately; changing the training setting does not perform deletion.
#1 Best Overall
- Model improvement: controlled by Improve the model for everyone or the equivalent Privacy Portal choice.
- Chat history: controlled by whether you use an ordinary or Temporary chat, and by deletion actions.
- Memory and personalization: controlled separately from model-improvement sharing.
- Third-party processing: controlled partly by the privacy policy of any service that receives information through a GPT action or plugin.
Turn off model improvement on a personal account
Signed-in Free, Plus, or Pro account
Use this path in the ChatGPT web or mobile interface:
- Open the account menu by selecting your profile name or account icon.
- Select Settings.
- Open Data controls.
- Turn Improve the model for everyone off.
- Select Done if the interface presents a confirmation button.
OpenAI says conversations started after the change will not be used to improve its models. Because this is an account setting, the choice carries to other devices where that same account is signed in. You do not need to repeat the change on every phone, tablet, or computer.
Signed-out use
If you use ChatGPT without signing in, open ChatGPT Settings and change the same model-improvement control. In this mode, the choice is saved in that browser. It does not follow you to another browser or device because there is no signed-in account connecting those sessions.
Free tools Windows power users keep installed
One-click scans. No signup required.
Clear cookies, private-browsing sessions, browser profiles, or managed-device policies can affect whether a signed-out preference remains available. Check the setting again before sending sensitive material from a browser where you are not signed in.
Use the Privacy Portal as an alternative
Personal Free, Plus, and Pro plans also offer a Do not train on my content choice in the Privacy Portal. OpenAI says either that choice or the in-product toggle is sufficient for ChatGPT conversations and Codex tasks. You do not need to submit both requests to opt out the same personal account.
The Privacy Portal and the Data controls page may not look identical, and availability can depend on plan, region, account state, or workspace. If you can access the in-product toggle, it is the most direct way to confirm the current setting for the account you are using.
Keep one conversation out of history with Temporary Chat
Use Temporary Chat when the goal is more than model-improvement opt-out. A temporary conversation does not appear in chat history, does not create or update memories, and is not used to improve OpenAI models while it remains temporary.
Starting a Temporary chat
- Start a new conversation.
- Select Temporary before sending the first message.
- Choose Personalized or Unpersonalized.
- Send your message only after the temporary mode and personalization choice are visible.
A Temporary chat is a per-conversation choice. It does not change the account-wide model-improvement toggle for your other chats.
Personalized versus Unpersonalized
Personalized Temporary Chat can use existing memories, custom instructions, and plugins to tailor its replies. It still does not create or update memories while the chat remains temporary.
Choose Unpersonalized when you do not want those existing personalization inputs used for that conversation. This is useful for a one-off question about a confidential game project, an unreleased mod, or any other subject where prior profile context could reveal more than you intended.
Temporary retention and saving
Temporary Chat is not the same as immediate destruction. OpenAI may retain a copy for up to 30 days for safety. That retention period is a maximum stated for safety purposes, not a promise that every temporary conversation is kept for exactly 30 days.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →If you save a Temporary chat, it becomes a regular chat. Once saved, it follows normal history, personalization, and model-improvement settings rather than the temporary treatment. Decide whether you need a permanent record before selecting any save or export action.
Check the mode before sharing
The point at which you choose a setting matters. The model-improvement toggle applies to new conversations, so changing it after an ordinary conversation has already been sent does not retroactively change that earlier conversation. Similarly, selecting Temporary after a message has been sent cannot turn the earlier message into a temporary one. Start a fresh chat and confirm the mode before entering confidential text, files, identifiers, or code.
For an ordinary chat, reopen Settings > Data controls and verify that Improve the model for everyone remains off. For a Temporary conversation, look for the Temporary label and the selected Personalized or Unpersonalized choice before sending. If you switch between accounts, browsers, or a personal and managed workspace, check again because the applicable policy and scope may differ.
How the main choices differ
| Choice | Appears in history | Memory behavior | Model improvement | Retention described by OpenAI | Scope |
|---|---|---|---|---|---|
| Ordinary chat with Improve the model for everyone off | Yes | Uses normal memory and personalization settings | New conversations are not used to improve models | Normal account chat-retention controls apply | Account-wide on signed-in devices; browser-scoped when signed out |
| Temporary, Personalized | No while temporary | May use existing memories, custom instructions, and plugins; does not create or update memories | Not used to improve models while temporary | Safety copy may be retained for up to 30 days | That conversation only |
| Temporary, Unpersonalized | No while temporary | Does not use the usual personalization inputs; does not create or update memories | Not used to improve models while temporary | Safety copy may be retained for up to 30 days | That conversation only |
The table shows why the two controls are complementary. The account toggle lets you keep ordinary history while stopping model-improvement use for new chats. Temporary Chat removes the conversation from history and prevents memory changes, but it is still subject to the stated safety-retention period and any third-party policy involved in the conversation.
Memory, personalization, and deletion are separate decisions
Turning off Memory or another personalization feature does not turn off model-improvement sharing. Likewise, turning off model improvement does not erase memories, hide existing chats, or prevent every limited use of context for safety. OpenAI describes rare, high-risk situations in which safety-related context may still be used even when Memory or personalization is disabled.
If your objective is to remove a remembered fact, use the memory controls rather than relying on the model-improvement toggle. If your objective is to remove a transcript from the account, delete the chat. If your objective is to prevent a new transcript from entering history and from changing memory, use Temporary Chat before you send the first message. These are different operations and may need to be performed separately.
Exceptions that can override the simple opt-out expectation
Submitting thumbs-up or thumbs-down feedback
On personal plans, if you submit a thumbs-up or thumbs-down rating after opting out, the entire conversation associated with that feedback may be used to train models. The feedback action therefore creates an exception to the ordinary expectation that a new conversation is excluded after the toggle is off.
For a sensitive conversation, avoid sending either rating after opting out unless you accept that the associated conversation may become eligible for training.
Recommended Free Tools
GPT actions, plugins, and other recipients
If a GPT uses an action or another integration that sends information to a third party, that recipient’s privacy policy governs the information it receives. The recipient may retain the data longer than Temporary Chat’s possible 30-day safety period. ChatGPT’s setting cannot rewrite the recipient’s retention schedule.
Before sending confidential game code, personal identifiers, unpublished design documents, or account details through an action, check what information the action transmits and which organization receives it. A Temporary chat can protect the ChatGPT-side history and memory behavior while the external recipient follows its own rules.
Safety processing
Temporary Chat can be retained for up to 30 days for safety, and disabling Memory does not remove limited safety-related context use in rare, high-risk cases. These are service-safety qualifications, not a way to restore model-improvement sharing after you opt out.
Personal accounts, managed workspaces, and API data
Personal Free, Plus, and Pro accounts may have model-improvement sharing enabled by default, but the account holder can opt out through Data controls or the Privacy Portal.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsFor Business, Enterprise, Edu, Healthcare, and API offerings, OpenAI says inputs and outputs are not used by default to improve models. That default does not mean an individual member controls every privacy setting. Organization owners and workspace administrators can set retention, access, memory, compliance, and related policies.
| Environment | Default model-improvement treatment | Who may control additional policies |
|---|---|---|
| Personal Free, Plus, or Pro | Sharing may be enabled by default; the user can opt out | The account holder controls the personal setting, subject to stated exceptions |
| Business, Enterprise, Edu, or Healthcare workspace | Inputs and outputs are not used by default to improve models | Workspace owners and administrators can control retention, access, memory, and compliance |
| API offering | Inputs and outputs are not used by default to improve models | Organization policies and account administrators govern access and retention |
Privacy Center is rolling out to signed-in Free, Go, Plus, Pro, and Business users on web and mobile. Enterprise, Edu, and Healthcare accounts are outside that rollout. Even when Privacy Center is unavailable, the underlying privacy settings remain in Settings; the controls shown can vary by plan, region, account, and workspace.
When the setting is missing or looks different
The exact menu can vary by plan, region, account state, device, and workspace. If Privacy Center is not displayed, use the existing controls in Settings. If you are in a Business, Enterprise, Edu, Healthcare, or other managed workspace, the administrator may have set policies that individual members cannot change. Confirm which account or workspace is active before assuming that a missing toggle means your preference was ignored.
A signed-out preference is browser-scoped, so a different browser, cleared cookies, private window, or another device may not retain it. A signed-in preference follows the account across signed-in devices, but it does not automatically govern a separate account. Recheck Data controls whenever you change accounts or workspaces.
Delete chats when removal is the goal
Opting out does not remove chats that are already in your history. To remove a saved conversation, use the chat’s delete command rather than changing Data controls. OpenAI says deletion removes the chat from the account view immediately and schedules permanent deletion within 30 days, unless legal or security obligations require longer retention.
Deleting a chat and turning off Memory are also separate. Removing a transcript does not make the model-improvement toggle unnecessary for future conversations, and changing the toggle does not delete a transcript that is already saved. A chat that was archived or placed in a project remains subject to the same separate deletion decision.
A practical checklist before sharing sensitive information
- For an ordinary chat you want to keep in history, confirm Settings > Data controls > Improve the model for everyone is off.
- If you are signed out, remember that the preference is saved only in that browser.
- For no history and no memory updates, start a new Temporary chat before the first message.
- Select Unpersonalized Temporary Chat when existing memories, custom instructions, or plugins should not supply context.
- Do not submit thumbs-up or thumbs-down feedback on a sensitive personal-plan conversation after opting out.
- Review any GPT action or plugin recipient’s privacy policy before sending information outside ChatGPT.
- Delete saved chats separately when account removal is required.
- In a work or school workspace, ask the administrator which retention, access, and compliance rules apply.
Which option should you use?
You want to keep your chat history
Turn off Improve the model for everyone. Ordinary chats remain available in history, and the setting applies to new conversations on the signed-in account.
You need a one-off private conversation
Start Temporary Chat and choose Unpersonalized if you also want to avoid existing personalization inputs. Do this before sending the first message; changing settings after a message has been sent cannot retroactively make that ordinary message temporary.
You already have a saved conversation you want removed
Delete the chat. The model-improvement toggle addresses future sharing behavior, not removal of an existing transcript.
You use ChatGPT through an employer, school, or API project
Check the workspace or organization policy. Those environments have different defaults, and an administrator may control retention or access regardless of an individual user’s personal preferences.
Common mistakes to avoid
- Confusing opt-out with deletion: Turning off training use leaves saved chats in place.
- Using ordinary chat for a no-history task: The account toggle does not hide a transcript. Use Temporary Chat instead.
- Choosing Personalized by default: Personalized Temporary Chat can draw on existing memories, custom instructions, and plugins. Choose Unpersonalized when that context is not wanted.
- Rating a sensitive conversation: Thumbs-up or thumbs-down feedback can make the associated conversation eligible for training on a personal plan.
- Forgetting external recipients: An action or plugin can send content under a third party’s policy and retention schedule.
- Assuming one browser setting covers every session: Signed-out preferences are saved only in the browser where you set them.
- Relying on a personal preference in a managed workspace: Administrators may control retention, access, memory, and compliance.
Frequently Asked Questions
Can I keep my ChatGPT history and still opt out of model training?
Yes. Turn off Settings > Data controls > Improve the model for everyone. New conversations will not be used to improve models, while ordinary chats can remain in your history.
Does turning off the toggle delete my old chats?
No. The opt-out affects new conversations and does not erase saved, archived, or project chats. Delete a chat separately when removal is required.
Does Temporary Chat guarantee that no copy is retained?
No. Temporary Chat stays out of history and is not used to improve models while temporary, but OpenAI may retain a copy for up to 30 days for safety.
Should I choose Personalized or Unpersonalized Temporary Chat?
Choose Personalized if existing memories, custom instructions, or plugins may help the conversation. Choose Unpersonalized when you do not want those existing personalization inputs used. Neither option creates or updates memories while temporary.
Does disabling Memory also disable model-improvement sharing?
No. Memory, personalization, model-improvement sharing, and deletion are separate controls.
Can feedback override my opt-out?
On personal plans, submitting thumbs-up or thumbs-down feedback after opting out may make the entire associated conversation eligible for model training.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What happens if a GPT action sends my information elsewhere?
The third-party recipient’s privacy policy governs the information it receives and may allow longer retention than ChatGPT’s Temporary Chat safety period.
Do Business, Enterprise, Edu, Healthcare, and API users need this opt-out?
OpenAI says inputs and outputs in those offerings are not used by default to improve models. Administrators can still set separate retention, access, memory, and compliance policies.
Why can’t I find Privacy Center?
Privacy Center is rolling out to signed-in Free, Go, Plus, Pro, and Business users on web and mobile. Enterprise, Edu, and Healthcare accounts are outside that rollout, and availability also varies by plan, region, account, and workspace. The underlying settings remain in Settings.
The Bottom Line
For a personal account, switch Improve the model for everyone off in Settings > Data controls to exclude new conversations from model improvement while keeping ordinary history. Use Temporary Chat before the first message when you also need no history and no memory updates. Delete existing chats separately, avoid feedback on sensitive opted-out conversations, and review administrator or third-party policies whenever your chat leaves a personal account.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
