Stop AI Companies Using Your Chats to Train Models

Updated 11 November 2026 · Standard formula, shown below

You can often opt out of having future conversations used to improve an AI model, but the setting depends on the service and account type. Review each service’s data controls, turn off model-improvement options, and avoid sharing sensitive information; an opt-out may not erase data already collected or affect every use of your information.

What You’ll Need

Access to the account and the app or website where you use the AI service.
A few minutes to review its privacy or data settings and, if needed, its privacy policy.
For work or school accounts, your administrator or organization’s privacy guidance; account controls can differ from personal accounts.

Step-by-Step

List the services and accounts you use. Include AI chat apps, assistants built into other products, and any third-party tools that send your prompts to an AI provider. Note whether each account is personal, work, or school. Opting out in one product does not automatically change settings in another product, even if they use the same provider.
Open the service’s privacy or data controls. Sign in through the official app or website. Look under names such as Settings, Privacy, Data Controls, Personalization, or Chat History. Menu names change, so use the settings search or the provider’s official help center if you cannot find the relevant section.
Turn off model improvement or training use. Look for a control stating that conversations may be used to improve, train, or develop models, and disable it. Read the nearby description: some switches apply only to future chats, while others may change whether conversations are saved or used for product improvement. Do not assume that turning off chat history also turns off training unless the service says so.
Check temporary, private, or no-history chat options. If the service offers a temporary chat mode, review its stated data handling and use it for conversations you do not want kept in ordinary history. Temporary does not necessarily mean that data is never retained; providers may keep it for a limited period for safety, abuse prevention, or other disclosed purposes.
Review feedback and sharing choices. A thumbs-up, thumbs-down, report, or feedback form may allow the provider to review the conversation, even when general model-improvement use is disabled. Avoid submitting feedback with sensitive content, and check the prompt shown before you send it. Also review options for sharing chats publicly or creating shared links, and remove links you no longer want accessible.
Check connected apps and third-party tools. If you use an AI feature inside another app, check that app’s privacy settings too. For browser extensions, automation tools, or services that connect to an AI account, review their data permissions and privacy notices. An opt-out at the AI provider may not cover data handled separately by another company.
Save the setting and confirm it persisted. Follow any Save, Confirm, or Turn off prompt. Reopen the page or return to the setting to verify it remains disabled. If the service sends a confirmation email or displays an account notice, keep it for reference. Repeat these checks on other accounts and devices where the setting is managed separately.
Request deletion if you want existing data removed. Find the service’s chat deletion, data export, or privacy request options. Delete conversations individually or clear history if those controls meet your needs. If you want broader account data deleted, use the provider’s official privacy request process. Deleting visible chats may not remove data already de-identified, retained for safety, or included in a completed training process; check the provider’s policy for its limits and timelines.
Set a safer sharing habit. Before sending a prompt, remove names, contact details, account numbers, confidential work material, and other identifying information unless sharing it is necessary and permitted. Use fictional or generalized examples where possible. An opt-out is not a guarantee that a service will never process the conversation to provide the response or meet other disclosed operational needs.

Common Mistakes

Assuming one opt-out covers every service. Settings are usually specific to a product, account, or workspace. Check each provider and any app that forwards your prompts.
Confusing chat history with training controls. Hiding or deleting a conversation is not always the same as preventing model-improvement use. Read the exact setting description.
Thinking an opt-out is retroactive. A control may apply only to new conversations. Ask the provider what happens to older chats and whether a separate deletion request is available.
Overlooking feedback and shared links. Feedback can trigger human review, and a shared link can expose a conversation to people who have it. Treat those actions as separate from training preferences.
Relying on a private browser window. Incognito or private browsing does not change an account’s data policy or prevent the service from processing prompts you submit while signed in.

Tips

Use a password manager or private checklist to record which accounts you reviewed and when. Recheck after major app updates, account changes, or policy notices.
For an employer or school account, ask the administrator what controls apply. An organization may manage retention and training settings centrally, and individual users may not be able to change them.
Read the provider’s current privacy notice, especially sections on retention, service improvement, human review, and legal or safety exceptions. Policies and controls can change.
When a setting is unclear, contact support with a specific question: whether future prompts are excluded from model training, whether past chats are affected, and what retention exceptions apply.

Troubleshooting

You cannot find a training switch: Search the settings page for “improve,” “training,” “data,” or “privacy.” Then consult the official help center for your account type. If there is no opt-out, do not assume that chat deletion or temporary mode is equivalent; limit sensitive prompts and ask support about available choices.

The control is unavailable or locked: You may be using a managed work or school account, a regional version, or a feature with different controls. Ask the administrator or provider which settings apply. Do not create another account to bypass organization rules.

The option turns back on or does not save: Check for a confirmation step, update the app, sign out and back in, and verify the setting on the service’s website. If it still changes, contact support and include the setting name and a screenshot with personal information hidden.

You deleted a chat but still see it elsewhere: Check archived chats, shared links, connected apps, and any separate history or activity page. Deletion may take time or have stated exceptions. Use the official privacy request route for unresolved requests, and keep the confirmation.

Frequently Asked Questions

Does turning off training stop a service from processing my prompt?
No. The service still needs to process a prompt to generate a response. The setting usually concerns whether conversations may be used for model improvement, not whether they are handled at all.

Does deleting a conversation remove it from a trained model?
Not necessarily. Deletion may remove the chat from your account but cannot always undo prior use in model development. Check the provider’s policy or ask support what deletion covers.

Is temporary chat completely private?
Not automatically. It may stay out of normal history or avoid some improvement uses, but providers can disclose limited retention or review for safety and operations. Read the mode’s description.

Do paid accounts automatically opt out?
Do not assume so. Plans and account types can have different terms, and settings can change. Check the controls and current policy for the specific account you use.

Can I opt out for a work or school account?
Sometimes, but the organization may control the relevant settings. Ask its administrator what data is retained and whether prompts are used to improve models.

What if a company offers no opt-out?
Use only non-sensitive, generalized prompts, avoid uploading confidential files, and consider another service only after checking its privacy terms and controls.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.