If you use Claude on a Free, Pro, or Max plan, your chats may be training their next model.
There's a toggle that turns it off, but most people don't know it exists.
On those plans, Anthropic's default is to use your conversations and coding sessions to improve their models. Everything you type goes to their servers and can end up in a training run. Client details, code, and the stuff you'd call IP if a competitor saw it.
But it only takes about 30 seconds to fix:
1. Log in at claude.ai
2. Click your name in the bottom left
3. Settings, then Privacy
4. Find "Help improve Claude" and toggle it off
That's it. Anything you type after that stays out of training. Anything already in a training run stays where it is, so do this today.
Note that this applies to personal plans only. Team, Enterprise, and API usage are excluded by default. So if you're on a company plan, just check with the plan administrator to make sure they didn't change the setting.