ChatGPT conversations are not public by default, but they should not be treated as an absolutely confidential vault. What happens to a conversation depends on your account type, data controls, retention rules, whether you share a link, and whether you connect an outside service. OpenAI provides settings that let personal-account users control whether new conversations are used to improve its models, and Temporary Chat offers a different history and personalization experience. Business and other managed workspaces have separate data commitments.

The practical answer is to understand several different questions that are often confused: Can other users see a normal chat? Can OpenAI use it to improve models? How long can a conversation be retained? Does turning off training delete old chats? And what happens when you share a conversation or send information to a third-party tool?

This guide explains the distinctions using OpenAI’s current documentation. Settings and product behavior can change, and some options vary by account, country, subscription, or workspace, so confirm the controls available in your own ChatGPT account before sharing sensitive information.

Are ChatGPT conversations private from other users?

A normal ChatGPT conversation is associated with your account and does not automatically become a public webpage that other ChatGPT users can browse. That does not mean every conversation is protected by the same rules as information held under a professional confidentiality agreement. OpenAI processes information to operate and secure its services, and its policies describe circumstances in which content may be reviewed or used.

The most common way a user deliberately exposes a conversation is by creating a shared link. OpenAI’s shared-links documentation explains that a personal-account link can be opened by anyone who has it. It does not offer recipient-by-recipient access controls or a configurable expiration date. A recipient may forward the link, and deleting the link later does not remove a copy another person already saved.

Before sharing, inspect the preview carefully. Depending on the sharing option and account, the shared content may include a conversation snapshot, assistant responses, supported images, or uploaded files. Do not assume a shared link reveals only the last answer; a link created from a conversation can include earlier messages up to the point it was shared.

You can review and remove links from Settings → Data controls → Shared links → Manage. Revoking a link prevents future access through that link, but it cannot erase copies already saved by other people. Also remember that changing model-training settings does not remove existing shared links.

Practical rule: keep ordinary conversations unshared unless you intend to distribute their contents. Before sending a link to a colleague or posting it publicly, read the full preview and remove personal details, private documents, and other material the recipient does not need.

Can OpenAI use your conversations to train its AI models?

For personal ChatGPT workspaces, OpenAI says content may be used to improve its models unless you opt out. The relevant setting is called Improve the model for everyone. When it is turned off, new conversations remain in your chat history but are not used to train OpenAI models, according to the Data Controls FAQ.

This is an important distinction: model training and chat history are separate controls. Turning off training does not automatically delete your conversations, and deleting a conversation is a separate action. If you want to keep your history while opting out of training, you can do so without switching to Temporary Chat.

To change the setting on the web or mobile app:

  1. Open the account or profile menu.
  2. Select Settings.
  3. Open Data controls.
  4. Turn off Improve the model for everyone.

OpenAI says this preference applies to your account across devices. The wording and placement of controls can change, so use the Data Controls section shown in your current version of ChatGPT.

The setting applies to new conversations after you opt out; it should not be interpreted as a promise that information already used in a past training process is retroactively removed. If you submit product feedback—such as a thumbs-up or thumbs-down—OpenAI’s documentation says the conversation associated with that feedback may be used to improve models even if you opted out. Avoid sending feedback on a conversation containing sensitive information if that consequence is not acceptable to you.

For details, read OpenAI’s explanation of how data is used to improve model performance. If you use other OpenAI products, do not assume that one setting controls every product: some services have separate settings. For example, OpenAI documents a separate environment-data control for Codex.

What does Temporary Chat actually do?

Temporary Chat is useful when you do not want a conversation to appear in your normal chat history. OpenAI says temporary conversations are not used to improve its models while they remain temporary, do not create or update memories, and do not appear in your chat history unless you choose to save them.

However, “temporary” does not mean that no copy can ever exist on OpenAI’s systems. According to the Temporary Chat help page, OpenAI may keep a copy for up to 30 days for safety purposes. The feature is therefore a different retention and personalization mode, not a guarantee of zero processing or instant deletion.

The current experience also allows a choice about personalization before starting a temporary conversation. If you choose a personalized temporary chat, it may use existing memories, custom instructions, and plugins to tailor responses, but it does not create or update memories while it remains temporary. If you choose an unpersonalized chat, those personalization sources are not used. Your account or workspace restrictions still take priority.

If you save a temporary conversation, it becomes a regular chat and follows your account’s normal personalization and model-improvement settings. That matters if you use Temporary Chat to discuss a one-off topic and later decide to keep the conversation.

To use it, start a new chat and select Temporary in the ChatGPT interface before sending the first message. The exact control may appear in a different place across app versions. If the option is not visible, consult the current help page rather than assuming that an ordinary chat is temporary.

Use Temporary Chat when you want a conversation kept out of your history and excluded from model improvement while it remains temporary. Do not use it as a place to paste passwords, full payment-card details, confidential business records, or other information that should not be disclosed to an online service.

How long does ChatGPT keep conversations?

There is no single retention period that applies to every type of ChatGPT data. Regular conversations generally remain in your account until you delete them or a relevant workspace retention policy applies. Temporary Chat follows a different process and may be retained for up to 30 days for safety purposes. Other data, such as shared links, uploaded files, or information handled by connected services, can have separate rules.

OpenAI’s consumer data controls guide describes the available options for managing conversations, model-improvement preferences, and other personal-data controls. Review the current policy for the product and account you use rather than assuming that deleting one chat deletes every associated item in every system.

For regular chats, deleting a conversation is different from archiving it. Archiving is a way to remove a chat from the main list without treating it as deleted. If you want content removed from your chat history, use the delete option and review OpenAI’s current retention explanation for what happens after deletion.

Shared links need separate attention. OpenAI says you can manage them from Data controls. Removing a shared link stops future access through that link, but it does not remove a recipient’s saved copy. Similarly, turning off model training does not delete conversation history or shared links.

Uploaded files and connected applications introduce additional considerations. A file can be handled by a particular product feature or stored in a product-specific library, while information sent through a third-party integration may be subject to that provider’s own terms. If your aim is to minimize the amount of information retained, check the controls for each feature you used—not only the conversation itself.

Are ChatGPT Free, Plus, and Pro conversations handled differently?

For personal workspaces, Free, Plus, and Pro users can manage whether new conversations are used to improve OpenAI models through the relevant data controls, where the setting is available. A paid subscription should not be interpreted as an automatic guarantee that personal conversations are excluded from model improvement. Check the current setting in your own account.

OpenAI treats business and managed workspaces differently. Its documentation states that content from ChatGPT Business, Enterprise, Edu, and certain other business offerings is not used to train models by default. Organizations may also have workspace-level retention, access, compliance, and administrative settings. Those controls and commitments depend on the specific product and agreement.

This is why it is important to distinguish the price of a subscription from its data terms. A consumer subscription can provide more model access or higher limits without necessarily providing the same organizational controls as a managed business workspace.

For an organization evaluating ChatGPT, the relevant questions include who administers the workspace, what retention policies apply, whether administrators can manage shared links, what contractual commitments are in place, and which data is permitted under company policy. Employees should not assume that using a work email address alone turns a personal account into a managed business workspace.

For more context on plan differences, see our guide to ChatGPT Free vs. Plus. Always read the terms for the specific plan and workspace before uploading confidential information.

What happens when you use a custom GPT, connector, or external tool?

A conversation can involve more than ChatGPT itself. If a custom GPT uses an action or a connected third-party service, information needed to complete the request may be sent to that service. OpenAI’s Temporary Chat documentation warns that data sent to a third party through a GPT action is governed by the recipient’s privacy policy and may be retained for longer or used for other purposes.

The same general caution applies whenever you connect an external account or authorize an integration. Before enabling it, identify which data the service can access, what information is sent in each request, and whether the provider’s terms permit the intended use. Only grant permissions that are necessary for the task.

A feature being built into the ChatGPT interface does not mean every participating service follows identical retention and privacy rules. Check the description of the specific connector or action, and avoid using it with confidential material unless you understand and accept the data flow.

If you are working with client documents, unpublished research, employee records, source code, or regulated data, follow your organization’s approved-tool policy. A personal account and a managed workspace may have different protections, and a third-party action may introduce a separate data recipient.

Can you delete or export your ChatGPT data?

ChatGPT provides controls to export data associated with your account and to delete conversations or your account. The available options can vary depending on whether you are using a personal account or a managed workspace. To review them, open Settings → Data controls and consult OpenAI’s current instructions.

Exporting data can help you keep a copy of your information or understand what is associated with your account. It does not, by itself, delete the data stored in the service. Deleting a chat is also separate from turning off model improvement. Choose the action that matches your goal.

If you want to delete shared links, manage them separately from the same Data controls area. If another person has already saved content from a shared link, revoking that link cannot remove the copy from the other person’s account. Avoid sharing sensitive content in the first place if you cannot control downstream copies.

Account deletion is a more consequential step than deleting an individual conversation. Read OpenAI’s account-deletion instructions carefully before proceeding, particularly if you need to export information or understand how the deletion affects access to other services.

What should you avoid sharing with ChatGPT?

Even with privacy controls enabled, the safest practice is data minimization: provide only the information required to complete the task. A privacy setting is not a substitute for judgment about what an online service needs to receive.

Avoid entering or uploading:

  • Passwords, recovery codes, API keys, private tokens, or full payment-card details.
  • Government identification numbers or financial records unless the service and workflow are specifically approved for them.
  • Confidential client information, unreleased business plans, or proprietary source code without authorization.
  • Personal information about other people that is not necessary for the task.
  • Medical, legal, or employment records that your organization’s rules prohibit sharing with consumer AI services.

When you need help with a sensitive document, consider removing names, account numbers, addresses, and other identifiers first. Replace them with consistent placeholders so the assistant can still follow the structure. For example, “Client A” and “Invoice B” may be sufficient for a formatting question that does not require real identities.

For a technical problem, provide the error message and a minimal code example rather than uploading an entire private repository. For a spreadsheet, share a small anonymized sample with the relevant columns. For a document summary, remove appendices or personal data that are not needed to answer the question.

These precautions reduce unnecessary exposure, but they do not guarantee that anonymized data can never be reidentified. Use the minimum data necessary and follow the policies that apply to your work.

A practical ChatGPT privacy checklist

Use this checklist when reviewing your settings:

  1. Check model improvement. Open Settings → Data controls and decide whether to turn off Improve the model for everyone for new conversations.
  2. Use Temporary Chat when appropriate. Confirm that the interface shows Temporary before sending a message, and remember that a copy may be kept for safety purposes for up to 30 days.
  3. Review shared links. Open Data controls → Shared links → Manage, and delete links you no longer want people to access.
  4. Check personalization. Review Memory and custom instructions separately; these settings are not the same as model-training controls.
  5. Audit connected tools. Check what information custom GPT actions and external services can receive.
  6. Minimize sensitive content. Remove identifiers, credentials, and confidential details that are not needed.
  7. Use the correct workspace. Confirm whether you are in a personal account or an organization-managed workspace before sharing work data.
  8. Recheck after product changes. OpenAI may update feature names, defaults, and controls, so revisit the official documentation when privacy requirements matter.

So, are ChatGPT conversations private?

The accurate answer is conditional. Ordinary chats are not automatically public, and OpenAI provides controls for model improvement, temporary conversations, data export, deletion, and shared-link management. But those controls serve different purposes. Turning off model improvement does not erase history; Temporary Chat can still be retained for safety purposes; a shared personal link can be viewed by anyone who has it; and information sent to an external service follows that service’s applicable terms.

For routine questions and low-risk tasks, the default experience may be sufficient for your needs. For sensitive work, use the appropriate workspace, review the settings, limit what you share, and verify every external connection. If your organization has a formal data-handling policy, follow it even when a feature appears convenient.

Privacy is not a single switch. It is the result of choosing the right account, understanding each control, and sharing only what is necessary.

Official sources