Take back control of your AI data
Your conversations with AI chatbots may be used to train future models. This guide walks you through how to opt out of data collection on every major AI platform — step by step. Use the jump links below to go straight to a platform.
⚠ Keep in mind: Privacy settings change frequently. These steps reflect policies as of February 2026. Always check each platform’s official privacy policy for the latest information, as defaults and menu locations can shift with product updates.
OpenAI
ChatGPT
Default setting: Data collection for model training is ON for free and paid personal accounts. Business and Enterprise users are excluded by default.
Steps to opt out — Web
- Sign in at chatgpt.com
- Click your profile icon in the bottom-left corner
- Select
Settings - Click
Data Controls - Find "Improve the model for everyone" and toggle it Off
💡 Tip: Use Temporary Chat (icon at the top of the chat screen) — these chats are never saved to history and are never used for training, regardless of your settings. Voice mode has a separate training toggle in the same Data Controls menu.
Alternatively: Visit privacy.openai.com and click "Do not train on my content" to opt out without logging in.
Alternatively: Visit privacy.openai.com and click "Do not train on my content" to opt out without logging in.
Google
Google Gemini
Default setting: "Keep Activity" is ON by default for personal Google accounts. Google retains all chats for up to 72 hours regardless of this setting. Google Workspace accounts (work/school) are excluded from training by default.
Steps to opt out — Web or Mobile
- Sign in and go to gemini.google.com
- Click Settings & help in the bottom-left corner
- Click the activity/clock icon at the top, or go directly to myaccount.google.com/data-and-privacy
- Find "Gemini Apps Activity" or "Keep Activity"
- Click Turn off and confirm
- (Optional) Click "Turn off and delete activity" to also remove past stored chats
Additional step — Gmail smart features
- In Gmail, open
Settings›See all settings - Find "Smart features:"
- Uncheck "Use Gmail, Chat, and Meet to personalize your experience in other Google products"
- Save changes
💡 Remember: This setting must be repeated for each Google account you use. If you use Gemini Live (voice), also disable "Improve Google services with your audio and Gemini Live recordings" in the same activity settings page.
Microsoft
Microsoft Copilot
Default setting: Model training is ON for personal Microsoft Account users. Users with organizational accounts (Microsoft 365 for work/school) and users under 18 are excluded by default. Conversations are stored for 18 months by default.
Steps to opt out — copilot.com (Web)
- Sign in at copilot.com
- Click your profile icon in the lower-left
- Select Settings
- Select your profile name, then navigate to
Privacy›Model Training - Toggle off both "Model training on text" and "Model training on voice"
Steps to opt out — Mobile App
- Open the Copilot app and tap the menu
- Tap your profile icon
- Select
Account›Privacy›Model training on text - Toggle it off
- Repeat for Model training on voice
💡 Note: Opting out applies to past, present, and future conversations. The change may take up to 30 days to fully apply across Microsoft's systems. You can keep personalization features on while having model training turned off.
Anthropic
Claude
Default setting: As of October 2025, Anthropic changed its policy — model training is now ON by default for Free, Pro, and Max personal plan users. This is a change from their previous privacy-by-default stance. Commercial accounts (Team, Enterprise, and API users) are not affected and remain excluded.
Steps to opt out — Web
- Sign in at claude.ai
- Click your profile icon or initials in the left sidebar
- Select
Settings - Click the
Privacytab - Find "Help improve Claude" and toggle it off
Steps to opt out — Mobile App
- Open the Claude app and tap your profile icon
- Go to
Settings›Privacy›Privacy Settings - Toggle "Help improve Claude" off
⚠ Important: If you opt out, conversations are retained for 30 days. If you opted in (or didn't act before the initial deadline), data may be retained for up to 5 years. Conversations you manually delete are not used for training, regardless of other settings.
Meta (Facebook, Instagram, WhatsApp)
Meta AI
Default setting: Meta AI is embedded in Facebook, Instagram, Messenger, and WhatsApp. There is no true off switch for US users. Meta uses your public posts and AI chat interactions for model training. EU, UK, and some other regions have formal opt-out rights under local privacy law. Private messages are not used for training (WhatsApp messages are end-to-end encrypted).
Best available option for US users — Submit an objection request
- Log in to Facebook or Instagram
- Go to
Settings & Privacy›Privacy Centre - Find "AI at Meta" under Privacy Topics
- Select "Submit an objection request"
- Enter your account email, optionally describe your concern, then Submit
EU / UK users — Formal opt-out
- Log in to Facebook or Instagram
- Go to
Settings & Privacy›Privacy Centre›How Meta uses information for generative AI - Click "Right to object" and complete the form
⚠ Important for US users: Meta reviews objection requests individually and may deny them. This is not a guaranteed opt-out — it is a request. The most reliable ways to limit your exposure are to make your posts private or avoid using the Meta AI chat feature entirely.
💡 WhatsApp tip: End-to-end encrypted personal messages cannot be used for training. However, any messages sent directly to the Meta AI feature within WhatsApp are not encrypted and may be used. Simply avoid initiating or responding to the Meta AI chat inside WhatsApp.