ChatGPT Privacy Center: Where Settings Live

ChatGPT Privacy Center is a new in-product hub that explains privacy options and links to the settings that manage them. It does not replace ChatGPT Settings, change any control when opened, delete data, or override workspace policies. OpenAI is rolling it out on web, iOS and Android, with availability varying by plan, region, account and workspace. Official September 21 release note.

ChatGPT Privacy Center: what changed

The practical change is navigation and explanation. Privacy Center groups information about memory, personalized ads, location, temporary chat, plugins and apps, multi-factor authentication, model improvement, data export and account deletion. A topic can explain the option and, where available, provide a Manage link to the existing setting. Temporary chat instead offers an action to begin a new temporary conversation. OpenAI’s Privacy Center guide.

That distinction prevents a common misunderstanding: viewing a privacy dashboard is not the same as changing a privacy choice. OpenAI says opening the center does not change settings or delete data. Users still need to open the relevant control, understand its scope and complete the action they intend.

The release-note summary lists signed-in Free, Go, Plus and Pro users. The detailed Help Center page also lists Business users, while explicitly excluding Enterprise, Edu and ChatGPT for Healthcare from this rollout. We treat the detailed availability page as the fuller current scope, not as evidence that every listed account already has access. For general plan context, see our GPT-6 access guide.

How to open ChatGPT Privacy Center

OpenAI documents different entry points by surface:

On the web, open the account menu, choose Help, then Privacy center.
On iOS, open Settings, then Privacy Center.
On Android, open Settings, then Privacy center.
In a desktop app, use Privacy center from Help if the item appears; it opens the center in a browser.

If ChatGPT Privacy Center is not showing, OpenAI says the rollout may not have reached that account or app version. The options displayed can also depend on plan, region, account and workspace settings. The documentation does not provide a force-enable step. Existing privacy controls remain available through Settings even when the hub is absent.

Our editorial recommendation is to record the platform, app version, plan and workspace before troubleshooting a missing item. Then check the regular Settings menu for the specific control you need. This separates a missing explanatory hub from a missing privacy control and makes a support report more precise.

What the three Privacy Center categories cover

The first category, Personalized chats & ads, covers memory, ad personalization and location. OpenAI notes that deleting a chat does not automatically delete a memory saved from it. Ads may appear for eligible Free and Go users; the help page says Business, Enterprise, Edu and ChatGPT for Healthcare do not have ads. Location controls appear when available.

The second category, Chat privacy, explains temporary chat and can start one. The third, Data use & access, brings together information about plugins and apps, multi-factor authentication, model improvement, export and account deletion. Organization policies can determine which connected apps and controls appear. Our multiple-account plugin guide provides related connection-management context, but app permissions remain separate from Privacy Center navigation.

These categories are an index, not a single master toggle. Memory affects personalization. Model-improvement settings govern whether eligible content can help train OpenAI models. Temporary chat changes the handling of one conversation while it remains temporary. Data export and account deletion are different actions again.

ChatGPT data controls still do the work

OpenAI’s Data Controls guide says Settings > Data controls remains the place to manage model improvement and other available options. For signed-in accounts, turning off Improve the model for everyone applies across devices. New conversations will not be used to train OpenAI models, but they can still appear in chat history.

Turning off model improvement does not delete or hide regular, archived or project chats. Archiving a conversation does not change whether it can be used for model training; the account’s model-improvement setting continues to apply. If the goal is removal rather than training preference, deletion is a separate action.

Memory and model training are also separate. Memory controls personalization, while the training control determines whether eligible content can help improve models. This is why a review should inspect both rather than treating either switch as a universal privacy mode.

For a personal ChatGPT plan, OpenAI says Improve the model for everyone also applies to Codex tasks. Codex has a separate Include environments setting for additional environment context, and changing the ChatGPT training preference does not change that separate setting.

Temporary chat, export and deletion boundaries

While a chat remains temporary, OpenAI says it does not appear in history, create or update memories, or contribute to model improvement; it may be retained for up to 30 days for safety purposes. A temporary chat can now be personalized with existing memories, custom instructions and plugins if the user chooses that mode before starting. Saving a temporary chat converts it into a regular chat governed by the account’s usual settings.

Self-service export and account deletion are available through settings for eligible signed-in Free, Go, Plus and Pro users. Business, Enterprise and ChatGPT for Healthcare do not offer self-service export in the documented flow; eligible Edu workspaces may allow it when an administrator enables it. Managed workspaces do not offer self-service account deletion.

The Privacy Portal is separate from Privacy Center. The portal handles privacy requests, while the in-product center explains settings and routes users toward relevant controls. Likewise, this site’s privacy page describes gpt-6.org’s practices; it cannot change an OpenAI account or substitute for ChatGPT settings.

A five-minute privacy settings review

The following is our editorial checklist, not a claim that we changed or tested settings on a reader’s account:

Open Privacy Center or the regular Settings menu and identify the exact account and workspace.
Review memory separately from Improve the model for everyone.
Check connected apps and permissions against the services you still use.
Review multi-factor authentication as an account-security control, not a data-deletion tool.
If using temporary chat, decide whether personalization should be on before starting.
If the goal is removal, distinguish deleting a chat, deleting a saved memory and deleting an account.
In a managed workspace, confirm which controls belong to the organization or workspace owner.

Record only the setting name and intended outcome; do not copy sensitive account information into an audit note. A visible toggle is configuration evidence, while an export receipt, deletion confirmation or support response is separate completion evidence.

The useful takeaway is simple: ChatGPT Privacy Center makes privacy choices easier to find and understand, but it does not collapse those choices into one switch. Select the control that matches the intended outcome, verify its scope, and handle managed-workspace limits through the organization that administers the account.