How to turn off AI training in ChatGPT, Gemini, Claude, Copilot and Perplexity

Who trains on your input by default?

As of 21 October 2026, based on what the providers state. Menu labels can change with app updates.

How to opt out, service by service

ChatGPT

1. Open your account menu (profile picture).
2. Select Settings, then Data controls.
3. Turn off Improve the model for everyone and select Done.

In the mobile app, go via the sidebar and profile icon to Data controls. Signed in, the setting applies to your account on every device. You can also choose "Do not train on my content" in OpenAI's privacy portal. For a one-off conversation, use Temporary Chat: it stays out of your history and is not used for training.

Google Gemini

1. Go to gemini.google.com and click Activity.
2. Near the top, click On.
3. Choose Turn off, or Turn off and delete activity to remove past chats as well.

With Keep Activity off, Google does not use your future chats to train its models unless you send feedback. On a work or school account, your administrator controls this setting. Gemini's temporary chats are not used for training either.

Claude

1. Open the menu via your name and select Settings.
2. Open Privacy.
3. Switch off the model improvement toggle.

This covers Claude Free, Pro and Max. Anthropic then stops using your stored chats in future training runs; models already trained are not affected. Incognito chats are never used for training, according to Anthropic.

Microsoft Copilot (personal account)

1. Older app on copilot.com: profile icon, profile name, then Privacy.
2. Windows or Mac app: profile icon, Settings, then Privacy.
3. Turn off Training on conversation activity and Training on voice conversations.

In the new Copilot app (since 18 August 2026), Microsoft says prompts, responses and file contents are not used to train foundation models. Optional feedback may still be used to improve Copilot.

Perplexity

1. Open Account settings.
2. Go to Preferences.
3. Turn off the AI data retention toggle.

The toggle is on by default for Free, Pro and Max. Opting out only covers data collected afterwards; Perplexity says training data already collected cannot be removed.

What opting out does not do

• Your input still reaches the provider. Every message is sent and processed, otherwise there would be no answer.
• Data is still stored. ChatGPT keeps chats in your history until you delete them; deleted chats are scheduled for permanent deletion within 30 days, and temporary chats may be kept for up to 30 days. Gemini keeps chats for up to 72 hours even with Keep Activity off. Claude removes deleted chats from its back-end within 30 days.
• People may read it. Google says that even with Keep Activity off it uses chats to protect the service, with help from human reviewers. Chats that have been reviewed are kept for up to three years, even if you delete your activity.
• Feedback overrides your choice. A thumbs up or down in ChatGPT means the whole conversation may be used for training despite your opt-out. Anthropic keeps feedback, with the conversation, for up to five years.
• It is not retroactive. Whatever has already gone into a model stays there.
• Safety checks continue. If Anthropic's systems flag a chat as breaking its usage policy, it may be kept for up to two years.

Turning training off is worth doing, but the bigger question remains: what belongs in an AI chat at all? Passwords, bank details and customer data stay out, opt-out or not. Our free test shows in your browser whether a text contains such details.

What about companies?

Business offerings such as ChatGPT Business, Google Workspace, Claude Team or Microsoft 365 Copilot do not train on input by default, according to the providers. When employees use personal accounts instead, everything depends on their own settings. More in our guide May employees enter customer data into ChatGPT?. This guide is not legal advice.

Frequently asked questions