How to Stop ChatGPT, Claude and Gemini From Training on Your Chats

You can stop new consumer chats from being used for model improvement by changing each chatbot’s privacy setting: turn off “Improve the model for everyone” in ChatGPT, turn off “Help Improve our AI models” in Claude, and turn off “Keep Activity” in Gemini. These controls are different from deleting chat history, disabling memory, or using a temporary conversation.

This guide shows the current steps for ChatGPT, Claude, and Gemini, explains what each setting does, and highlights the privacy limits that still apply after you opt out.

Quick privacy settings for ChatGPT, Claude and Gemini

Chatbot	Setting path	Control to change
ChatGPT	Profile → Settings → Data Controls	Turn off “Improve the model for everyone”
Claude	Account menu → Settings → Privacy	Turn off “Help Improve our AI models”
Gemini	Settings & help → Activity	Turn off “Keep Activity”

Menu names can vary slightly between the web and mobile apps, but the underlying controls are described in the companies’ official privacy documentation.

How to stop ChatGPT from training on your chats

On the ChatGPT website

1. Open ChatGPT and select your profile icon.
2. Select Settings.
3. Open Data Controls.
4. Turn off Improve the model for everyone.

According to OpenAI’s current Data Controls FAQ, your conversations can remain in your chat history after this setting is disabled, but new chats will not be used to train ChatGPT.

On the ChatGPT mobile app

Open the account menu, select Settings, choose Data Controls, and switch off Improve the model for everyone. OpenAI says the preference applies to the account, so changing it on one device also affects the account on other devices.

Use Temporary Chat for a separate conversation

If you want a conversation that does not appear in normal history or use memory, start a Temporary Chat. OpenAI’s Temporary Chat FAQ states that temporary conversations are not used to improve its models while they remain temporary.

Temporary Chat and the training toggle solve different problems:

• Training toggle off: Normal conversations can stay in your history, but new chats are excluded from model training.
• Temporary Chat: The individual conversation stays outside normal history and does not create or use memories.

Do not assume Temporary Chat makes it safe to paste passwords, API keys, private client files, or other secrets.

How to stop Claude from using chats for model improvement

On Claude web

1. Open your account menu.
2. Select Settings.
3. Open Privacy.
4. Find Help Improve our AI models.
5. Turn the setting off.

On the Claude mobile app

Anthropic’s model improvement settings guide gives the same general path on mobile: select your name, open Settings, choose Privacy, and switch off Help Improve our AI models.

Anthropic says that when a consumer user turns off model improvement, previous and new chats or coding sessions will not be used for future model training. However, data already included in a training run or an already-trained model cannot be removed retroactively.

Consumer Claude and commercial Claude use different rules

The setting above applies to consumer offerings such as Claude Free, Pro, and Max. Anthropic’s commercial privacy documentation says inputs and outputs from products such as Claude for Work and the Anthropic API are not used for model training by default unless the customer opts into a specific program or explicitly submits material as feedback.

How to stop Gemini from using future chats to improve AI

1. Open Gemini.
2. Open Settings & help.
3. Select Activity.
4. Turn off Keep Activity.

Google’s updated Gemini Apps Privacy Hub says turning off Keep Activity stops future chats from being reviewed to improve Google services. Google also advises users not to enter confidential information they would not want a reviewer to see while the setting is enabled.

Use Gemini Temporary Chat

Gemini also offers temporary chats. Google says temporary conversations are not used to improve Google AI with help from human reviewers. They are useful when you do not want a specific conversation in normal activity, but they are not a substitute for careful data handling.

Work and school Google accounts

Data treatment can differ for managed Workspace accounts. Google states that, for eligible work or school accounts, chats and uploaded files in Gemini Apps are not reviewed by human reviewers or used to improve generative AI models. Your organization’s administrator may still control access, retention, and account settings.

Turning off training is not the same as deleting your chats

These actions are often confused:

• Opting out of training controls whether eligible conversations are used to improve future models.
• Deleting a chat removes it from your visible history and starts the service’s deletion process.
• Turning off memory controls whether the chatbot remembers information for personalization.
• Using Temporary Chat changes how one conversation is stored and personalized.
• Deleting an account is a broader action covering the account and associated data.

If your goal is to reduce exposure, turn off model improvement, review stored activity, delete sensitive old conversations, and disable memory features you do not want.

Does opting out protect old conversations?

The answer depends on the service and timing.

• ChatGPT: OpenAI describes the toggle as preventing new conversations from being used for training. Review old chats separately and delete anything sensitive.
• Claude: Anthropic says switching model improvement off prevents previous and new chats from being used for future training, but it cannot undo training already completed or currently in progress.
• Gemini: Turning off Keep Activity affects future chats. You can review and delete existing Gemini activity separately.

Because privacy policies and product controls can change, check the official page linked in each section rather than relying on an old screenshot or social-media post.

Can feedback still be used?

Feedback buttons can create a separate exception. When you submit a thumbs-up, thumbs-down, report, or support request, the service may receive the associated conversation so staff can investigate the issue or improve the product.

Before sending feedback, review the attached content and remove personal, confidential, or proprietary information when possible.

What you should never paste into an AI chatbot

Privacy controls reduce some data use, but they do not turn a cloud chatbot into a password manager or secure document vault. Avoid entering:

• Passwords, recovery codes, private keys, or API secrets
• Full payment-card or bank-account details
• Government identification numbers
• Private medical records with identifying information
• Unreleased company financials or confidential contracts
• Client databases and personal customer information
• Source code or documents your employer prohibits uploading

Replace real names, account numbers, email addresses, and unique identifiers with neutral placeholders. Share only the minimum information needed for the task.

Extra steps for safer chatbot use

Check connected apps and extensions

A chatbot may be connected to cloud storage, email, a browser, or workplace tools. Remove integrations you no longer use and limit permissions to the smallest practical scope. Before installing an AI add-on, follow our guide to checking whether a browser extension is safe.

Secure the network and account

Use a unique password and multi-factor authentication. When working away from home, verify the network before opening sensitive accounts; our guide explains how to tell whether public Wi-Fi is safe.

Use approved business plans for work data

If you handle company or client information, use the organization’s approved enterprise or business workspace rather than a personal account. Managed plans often have different training defaults, access controls, and administrative policies.

Consider account-free tools for low-risk questions

For simple prompts that do not need history or personalization, our list of AI tools that work without login may be useful. An account-free interface is not automatically private, so check its retention and training policy first.

Frequently asked questions

Will ChatGPT still save chats if I turn off model training?

Yes. OpenAI says normal conversations can remain in your history after you turn off Improve the model for everyone. Use Temporary Chat or delete individual conversations if you do not want them in normal history.

Does turning off Claude model improvement delete my history?

No. The setting controls model improvement, not visible chat history. Delete conversations separately if you no longer want them stored in your account history.

Can I use Gemini after turning off Keep Activity?

Yes, but some activity-based features and personalization may work differently. Google may also retain limited data for service, security, or legal purposes under its current policy.

Is Temporary Chat completely anonymous?

No. Temporary means the conversation receives different history, memory, and training treatment; it does not mean the provider cannot process limited technical or safety data. Do not share secrets simply because the temporary option is enabled.

Do business chatbot accounts train on company data?

Major providers generally publish stronger default protections for eligible business, enterprise, education, and API products, but the exact contract and administrator settings matter. Check your organization’s plan and data-processing terms.

Bottom line

To stop future consumer chats from training AI models, disable the model-improvement setting in ChatGPT and Claude, and disable Keep Activity in Gemini. Then review old conversations, memory, connected apps, and feedback settings separately.

The safest rule remains simple: even after opting out, never upload information that would cause serious harm if it were exposed, retained, or viewed by another person.

Featured image: Dan Nelson on Pexels.