How to Make ChatGPT Private: Settings and Practical Limits
Reduce unnecessary disclosure by checking training, memory, sharing and saved content separately. No single setting makes every ChatGPT workflow private.
Updated 5 min read
To make ChatGPT more private, review model-training settings, memory, temporary chats, shared links and connected access separately, then limit what you submit. There is no single setting that removes all storage, access or disclosure risks. For work, use only the account and information your organization has approved.
A team also needs a common answer to “what may we put here?” Tools like fenbs can keep the approved AI-use rules available to connected assistants, without pretending to change ChatGPT’s settings for you.
What should you change first?
- Confirm the active account and workspace. Use the organization-approved workspace for work and ask an administrator which settings it manages.
- Open Settings, then Data controls. If Improve the model for everyone is available and you do not want new conversations used for training, turn it off.
- Open Settings, then Personalization and Memory. Review the controls and remembered information available to your account, rather than assuming the training toggle covers them.
- Choose temporary chat for a conversation you do not want in regular history. Review its personalization choice before sending anything.
- Review existing shared links and disconnect apps you no longer need. These are separate disclosure routes.
- Remove unwanted chats and separately saved files where appropriate, following your organization’s retention rules. Then check what information your next prompt actually needs.
OpenAI’s data controls guide (opens in a new tab) gives the current training-setting instructions. Turning off Improve the model for everyone concerns new conversations; it does not delete saved chats. Follow the guide for your account rather than looking for a universal “private mode” switch.
How do training and memory differ?
Training concerns using eligible content to improve models. Memory concerns information available to personalize your responses. Choose the settings for each purpose, and do not use the fact that ChatGPT mentions something you previously supplied as proof of how training works.
OpenAI’s memory documentation (opens in a new tab) explains that turning memory off does not delete past chats, and deleting a chat does not necessarily remove a separate saved memory. Review remembered information and its sources when removing a detail. Available controls vary with the account and workspace.
Suppose you previously supplied a client’s name and now want to stop using it. Check where it appears rather than relying on a new prompt that says “forget that.” The original conversation, a separate memory and a saved file are different places to review; avoid repeating the name while testing the cleanup.
When does temporary chat help?
OpenAI’s temporary chat guide (opens in a new tab) says a chat stays out of history and is not used to improve models while it remains temporary. It can use existing memories, custom instructions and plugins when personalized; choose Unpersonalized before starting if you do not want those personalization sources used.
Neither choice creates or updates memories while the chat remains temporary. OpenAI may retain a copy for safety purposes, and saving the conversation converts it to a regular chat governed by the relevant account settings. Read the current guide before treating temporary chat as a promise that nothing is retained.
Use it for an approved one-off question rather than as permission to upload a prohibited document. A fictional example can often answer the same question. The narrower guide to using ChatGPT incognito covers that conversation choice in more detail.
Where can sharing and connected access expose information?
Review shared links even if you have changed training or memory settings. OpenAI’s conversation-sharing guidance (opens in a new tab) explains that personal-account links can be opened by anyone who has the link, while managed-workspace links have a different audience. Preview the shared content and remove links you no longer need.
Do not assume that a response-only share and a conversation share reveal the same material. Inspect the preview for names, source excerpts, images and attachments. Once a recipient has saved information, removing the original link is not a way to recall every copy.
Connected apps deserve the same deliberate review: identify the app, the access you approve and the information needed for the task. Check its terms and workplace approval before connecting it. Disconnect unused access, and remember that stopping future access is a different action from removing content already copied into a conversation.
A practical team rule might allow public research summaries but require approval before connecting an internal document source. Record the rule and the person authorized to approve changes in fenbs Decisions and rules, so a new assistant session has the same boundary.
What must you delete or keep separately?
Follow OpenAI’s chat and file retention guidance (opens in a new tab) for the content you want removed. Archiving keeps a conversation; it is not deletion. Files saved separately in Library can remain after the conversation is deleted, so review those copies as well.
Deletion follows the service’s retention process and can have security or legal exceptions. For organizational records, ask the responsible person before removing material subject to a retention requirement. Keep an approved conclusion in the proper work system rather than keeping an entire chat merely because one paragraph is useful.
| Action | Purpose | Separate check |
|---|---|---|
| Disable eligible model improvement | Limit training use of new conversations | Saved history and memory |
| Review memory controls | Manage personalization sources | Original chats and saved files |
| Remove a shared link | Stop future access through that link | Copies recipients already saved |
| Delete an unwanted conversation | Request removal of that chat | Library files and retention obligations |
What changes when your employer manages the workspace?
Your organization may control access, features and retention. Review the commitments for the exact product on OpenAI’s enterprise privacy page (opens in a new tab) and ask your administrator how they apply to your work. A personal preference does not override the organization’s policy or make a managed account private from every authorized organizational process.
Before starting a shared workflow, agree on the permitted account, permitted content, reviewer and cleanup responsibilities. Write a concrete instruction such as “Use approved public product material only; ask the project owner before uploading client files.” Avoid broad instructions like “be careful,” which leave each person to invent the boundary.
Keep the approved boundary with the work
Use fenbs to record the account and disclosure rules a person approved, then add tasks for settings checks and unresolved access questions. Connected assistants can read those rules before working and save the next steps afterward. Keep the record limited to approved instructions rather than the sensitive material those instructions protect.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
How to do this with fenbs
- Create a free account and add a project for your approved AI-use rules. One person can start free, with no credit card.
- Record the permitted information, approved workspace and exception process in Decisions and rules, identifying the person who made the decision.
- Create tasks to review training settings, memory, shared links and saved files. Record what was checked and what still needs an administrator.
- Use Integrations to connect a supported assistant. Ask it to read the rules at the start and save progress before it stops.
fenbs does not operate ChatGPT privacy controls or remove information from its systems. It is not the right fit if you need a tool that enforces those settings automatically; it keeps the approved working rules and follow-up tasks.
For another provider, see how to make Claude more private. For account and tool choices, use ChatGPT for business and Claude vs ChatGPT; the AI governance hub covers the wider review process.