ChatGPT now lets you turn off training on your chats

If you have ever hesitated to paste sensitive code or work notes into ChatGPT out of fear that it might train future models, OpenAI is finally addressing that anxiety.

OpenAI just rolled out a new privacy setting that allows any user to turn off chat history. Once disabled, new conversations will not be used to train or improve models, and they will disappear from the sidebar. OpenAI will still hold onto those chats for 30 days strictly to monitor for abuse before deleting them permanently. Alongside this toggle, a new export feature lets you download your entire conversation history via email.

Dedicated business tier incoming

For teams that need tighter controls, OpenAI also announced ChatGPT Business, a forthcoming subscription planned for the coming months. It will adhere to the API data usage policy by default, ensuring enterprise inputs are never fed back into the training pipeline.

Why it matters

We think this is OpenAI's direct response to mounting scrutiny from global privacy regulators, especially in Europe. Giving users a one-click opt-out is far smarter than hiding a data-exclusion request behind confusing forms.

For everyday professionals, it means you can finally test sensitive queries without worrying about your company secrets leaking into public datasets. The obvious downside is friction: you have to sacrifice the convenience of your chat history sidebar just to protect your privacy. Even so, teasing an enterprise-grade tier shows OpenAI is determined to win over cautious corporate clients who have kept their checkbooks shut until now.