🔒 Stop ChatGPT, Claude and Gemini from using your chats for AI training

September 17, 2026

By Kim Komando

⏱️ The 60-second setup

Keep your ChatGPT history: Open Settings from your profile menu > Choose Data Controls > Turn off Improve the model for everyone.

The setting applies across web, iPhone and Android when you’re signed into the same account. New conversations will stay in your history but won’t be used to train OpenAI’s models.

Protect your privacy: Go to OpenAI’s Privacy Portal > Choose Make a Privacy Request > Select the request to stop your content from being used for training > Verify your account.

This gets you to the same place without changing chat history. It’s a privacy request, not a secret feature.

ChatGPT Temporary Chat: Start a new chat > Select Temporary at the top of the conversation > Confirm Temporary is visible before adding sensitive information.

Temporary Chats don’t appear in history, create memories or train ChatGPT. OpenAI may retain a copy for up to 30 days for safety. That’s considerably less dramatic than instant deletion, but it does stay in the system temporarily. 

Claude has incognito chats, and Gemini offers temporary ones, too. Same basic idea: less memory, less training, fewer digital breadcrumbs following you around.

Claude: Open Settings > Choose Privacy > Turn off Help improve Claude.

This controls whether new consumer chats can be used for model improvement. Your chat history won’t automatically disappear. Claude Team, Enterprise and API data follow separate commercial policies and aren’t used for training by default.

Turning off Help Improve Claude also drops Anthropic’s retention on new chats from five years to 30 days.

Google Gemini: Open Gemini Apps Activity > Choose Turn off > Select whether you want to delete saved activity.

With activity off, chats aren’t used to train Google’s AI models unless you submit feedback. 

Typing a privacy request into an ordinary chat doesn’t change how that chat is handled. The mode or account setting does the work. The chatbot can’t pinky-swear its way around company policy.

⚠️ The catch: Opting out protects future conversations. It can’t pull information out of a model that was already trained, so get on this ASAP. 

“Not used for training” doesn’t mean “sealed forever in a titanium vault beneath Switzerland.” Safety, support, abuse and legal exceptions can still apply. Google says some conversations reviewed by humans can be kept separately for up to three years. 

Keep health records, financial details, passwords and confidential work documents out of consumer chatbots whenever possible, even after flipping every privacy switch available.