Stop your chats and posts being used to train AI
Most AI apps can learn from what you type unless you switch it off. These are the switches, app by app, as of October 2026. They move around, so if one's gone, search the app's help for the setting's name.
The steps
- 5 min
Step 1: ChatGPT: turn off “Improve the model for everyone”
Settings → Data controls. On the phone it's under your profile in the sidebar. It follows your account to every device. Rating an answer thumbs up or down can still send that chat for training.
- 5 min
Step 2: Claude: turn off the model-improvement setting
Settings → Privacy → “Help improve our AI models”. It stops new chats being used; incognito chats never are.
- 5 min
Step 3: Gemini: turn off Keep Activity
gemini.google.com → Settings & help → Activity → Turn off. Chats are still kept for 72 hours but aren't used for training unless you send feedback. Work and school accounts are up to the admin.
- 5 min
Step 4: LinkedIn: turn off Data for Generative AI Improvement
The link goes straight to the switch. It stops future use, not training that's already happened.
- 5 min
Step 5: X: untick the Grok training box
Settings → Privacy & Safety → Data sharing and personalization → Grok & Third-party Collaborators. The separate Grok app has its own: Settings → Data Controls → Improve the model.
- 10 min
Step 6: Meta: see whether you can object
There's no simple switch. Public posts and your chats with Meta's AI can be used; in some countries you can file an objection from your account. Private messages aren't used unless someone in the chat shares them with the AI.
- 15 min
Step 7: Do the kids' accounts too, and re-check in six months
These settings get renamed and moved between updates. A quick look twice a year is plenty.
Written by Helg and community members, not by any government. Check the official sites before you pay for anything.
General information, not legal or professional advice. Fees and rules change — check the linked official pages.