Privacy · don't learn from me
Training opt-outs
30-second gist~30s read
By default, several AI services use your free conversations to improve their next model. You can usually turn this off in one or two clicks. The opt-out doesn't make the AI worse for you — just removes you from the training pool.
If you ever paste anything you wouldn't want strangers to see, switch this off first.
If you want more
Where to look (the path may shift slightly)
- ChatGPT: Settings → Data Controls → "Improve the model for everyone" → off.
- Gemini: myactivity.google.com → Gemini Apps Activity → off.
- Copilot (consumer): Settings → Privacy → Conversation training → off.
- Claude: Off by default. No action needed.
- Perplexity: Settings → Account → AI Data Retention → off.
- Most "AI on top of ChatGPT" apps: they connect to OpenAI behind the scenes, and that pipe (called an API — the technical connection between two apps) doesn't train on what passes through it. So you're already opted out by default — but the app itself may keep its own copy. Check the app's privacy page.
What this doesn't fix
The opt-out stops future conversations being used to train. It doesn't pull your past data out of an already-trained model. It also doesn't prevent the company from keeping your conversations for safety/legal/abuse-prevention reasons — usually for 30 days, sometimes longer.