How can I make sure my ChatGPT chats are private?

Your ChatGPT chats are more private than you might think, but less private than you probably want. By default, OpenAI stores your conversations, may use them to train its models, and retains data from connected third-party apps. The good news is that you have real controls available, and most of them take less than a minute to set up.

This guide walks through every meaningful privacy setting in ChatGPT, from turning off chat history to protecting sensitive business conversations. Follow the steps in order for the most complete setup, or jump to the section most relevant to your situation.

Turn off ChatGPT chat history in your settings

Disabling chat history is the fastest single action you can take to stop new conversations from being stored in your account and used to train OpenAI’s models. When this setting is off, your new chats will not appear in the sidebar history and will not be used for model training going forward.

1. Click your profile icon in the top-right corner of ChatGPT.
2. Select Settings from the dropdown menu.
3. Navigate to Data Controls.
4. Toggle off Chat History & Training.

Once toggled off, the setting applies account-wide across all your devices immediately. According to OpenAI’s data controls announcement, conversations created after you disable this setting are retained for about 30 days for abuse monitoring only, then permanently deleted. You will not see them in your sidebar, and they will not feed into future model training.

One nuance worth knowing: the “Chat History & Training” toggle does two things at once. If you want to keep your history visible but simply stop your data from training the model, use the separate Improve the model for everyone toggle in the same Data Controls menu instead. ChatGPT also has a Memory feature under Settings > Personalization that stores specific facts across sessions. Turning off chat history does not automatically disable memory, so check that setting separately if you want a clean slate.

Delete existing chats and clear your conversation log

Turning off chat history stops new conversations from being saved, but your existing conversation log stays in place until you remove it. Deleting old chats reduces the amount of personal data sitting in your account.

Delete a single chat

1. On desktop, hover over the conversation in the left sidebar.
2. Click the three-dot menu (⋯) that appears.
3. Select Delete and confirm.

On Android, open the menu using the two lines in the top-left corner, press and hold the conversation title, then tap Delete in red at the bottom of the screen.

Delete all chats at once

1. Go to Settings > Data Controls.
2. Select Delete all chats.
3. Confirm the action. This cannot be undone.

Deleted chats are scheduled for permanent removal from OpenAI’s systems within 30 days, unless OpenAI is required to retain them for security or legal reasons. Per the OpenAI Help Center on deleting chats, deleted conversations cannot be recovered through the interface, the API, or by contacting support.

Two things to keep in mind after deleting. First, deleting a chat does not remove memories that ChatGPT saved from that conversation. To clear those, go to Settings > Personalization > Manage Memories and delete them manually. Second, if you previously shared a chat link and someone imported it into their own account, your deletion does not remove their copy. Manage your shared links separately via Settings > Data Controls > Shared Links, since shared links can be indexed by search engines.

Use a temporary chat for one-off private sessions

Temporary Chat is ChatGPT’s built-in incognito mode. Use it whenever you want to have a conversation that leaves no trace in your account history and does not contribute to model training.

1. Open a new chat in ChatGPT.
2. Click the pill-shaped Temporary button in the top-right corner of the page.
3. Start your conversation. On the macOS app, click the model name at the top of the window and select the Temporary Chat toggle instead.

Temporary Chat works exactly as you would expect: the conversation does not appear in your history, ChatGPT does not update its memory based on what you share, and the content is not used to train the model. OpenAI may still retain a copy for up to 30 days for safety monitoring, but that copy is not accessible through your account.

As of January 2026, OpenAI updated Temporary Chat to optionally carry your personalization preferences (memory, style, and tone) into the session while still keeping the conversation itself unsaved. If you want a true blank slate with no personalization, toggle that option off in the Temporary Chat settings. Your plan limits for messages and model access still apply in Temporary Chat. It only changes how your content is handled, not what features you can use.

One important caveat: if you use a Custom GPT that connects to a third-party API during a Temporary Chat session, the data you share may be governed by that third party’s privacy policy, not just OpenAI’s. Temporary mode only covers what happens within OpenAI’s platform.

Opt out of OpenAI’s model training on your data

By default, conversations on Free, Plus, and Pro personal accounts may be used to train OpenAI’s models. Opting out is straightforward and applies immediately across your entire account.

1. Go to Settings > Data Controls.
2. Find the Improve the model for everyone toggle.
3. Switch it off.

The opt-out is forward-looking. New conversations will not be used for training, but data already used in completed training runs cannot be retroactively removed. If you are on a ChatGPT Business, Enterprise, or Education plan, this setting does not apply to you because those plans exclude user data from model training by default.

There is one exception to watch for. According to OpenAI’s data usage documentation, if you give feedback on a response using the thumbs up or thumbs down buttons, the entire conversation associated with that feedback may still be used for training, even with the toggle off. If you have opted out of training, avoid using the feedback buttons. Also note that if you use ChatGPT while signed out, your conversations are used for training by default and you cannot access the opt-out toggle without logging in.

Free-tier users in the US should also check their ad controls. ChatGPT memories can be used for ad personalization on the free plan, and this is on by default. Look for the Ad Controls section in your settings. This is a limited rollout and may not be visible to all users yet.

Manage connected apps and API data access

Connected apps extend ChatGPT’s reach into your external accounts, including Google Drive, Gmail, SharePoint, Dropbox, and GitHub. Each connection is a potential privacy exposure point because those third-party services operate under their own terms, not just OpenAI’s policy.

1. Go to Settings in ChatGPT.
2. Select Apps (or Connected Apps on a browser).
3. Review the list of connected services.
4. Disconnect any app you no longer actively use by clicking the disconnect option next to it.

Disconnecting an app immediately cuts off ChatGPT’s access to new data from that service. It does not delete past conversations that already used data from that app. As noted by Privacy International’s ChatGPT guide, reviewing connected apps regularly is especially important for Plus and Pro users, where ChatGPT’s agent mode can interact with those apps with limited user intervention.

Connected apps are only available on paid plans (Plus, Business, Enterprise, and Education). Free users cannot connect apps. If you are on a Free, Plus, or Pro plan, the data access is all-or-nothing: there are no granular file permissions. To reduce exposure, make sure the Improve the model for everyone toggle is off, as described in the previous section, so that data pulled from connected apps is not fed into model training.

Also manage your shared chat links. Go to Settings > Data Controls > Shared Links to see any conversations you have made public and delete links you no longer want accessible.

Extra precautions for sensitive business conversations

The settings above significantly reduce your privacy exposure on a personal account. For business use involving confidential data, client information, or regulated content, you need to go further.

Choose the right plan for your risk level

Consumer plans (Free, Plus, Pro) may use your inputs to train OpenAI’s models by default. ChatGPT Business and ChatGPT Enterprise exclude business data from training by default. Enterprise additionally includes SOC 2 Type 2 certification, AES-256 encryption at rest, TLS 1.2+ encryption in transit, SAML SSO, and audit trail access. Enterprise Key Management (EKM) allows Enterprise workspaces to manage their own encryption keys via AWS, Google Cloud, or Azure: if a key is revoked, OpenAI can no longer read the stored data.

Understand what ChatGPT cannot protect

Even with every privacy setting enabled, ChatGPT is not a zero-knowledge vault. OpenAI staff and authorized contractors can still review flagged conversations for safety violations. Conversations are not end-to-end encrypted in the way a secure messaging app is. OpenAI can also be compelled to share conversation data under court order or government authority.

Conversations with ChatGPT carry no legal privilege. Unlike communications with a doctor or lawyer, sensitive legal or medical details typed into the chat could potentially be used against you in a legal proceeding. The Samsung incident from 2023 remains a widely cited example of what can go wrong: employees pasted confidential source code into ChatGPT, which was then processed by OpenAI and potentially used in model training, prompting Samsung to restrict employee access entirely.

Practical rules for business use

• Never paste proprietary source code, financial projections, or client data into a consumer ChatGPT account.
• If your organization operates under GDPR, CCPA, HIPAA, or financial regulations like DORA, use ChatGPT Enterprise or Healthcare plans that support Data Processing Addenda and Business Associate Agreements.
• If you use ChatGPT Enterprise’s agent mode, be aware that screenshots taken by the agent are retained for up to 90 days, longer than the standard 30-day retention for text conversations. Avoid having sensitive documents open in the same browser tab the agent is using.
• Establish a clear internal policy on what employees can and cannot share with any AI tool, regardless of plan tier.

For businesses that want their content to appear in AI-generated answers rather than just managing what they share with AI tools, AI visibility optimization is a separate but related discipline worth understanding. The same generative engines that process your ChatGPT queries also determine which businesses get cited in AI Overviews and ChatGPT responses to customer questions.

Privacy and visibility are two sides of the same AI landscape. Controlling what you share with ChatGPT protects your data. Structuring your public content correctly ensures your business gets found through those same systems. Both matter as AI becomes the primary way people discover information online.