ChatGPT Finally Gets an Off Switch for Your Chat History

OpenAI just made it a whole lot simpler to tell ChatGPT to forget what you said. The company is rolling out a new toggle that lets users turn off chat history entirely, meaning any conversation started with the setting disabled won’t be used to train its models or even show up in your sidebar. It’s a direct, overdue response to the privacy headaches and regulatory pressure that have dogged the chatbot since its meteoric rise. For a service that ingests everything from casual banter to sensitive work documents, a straightforward ‘don’t train on me’ button feels like it should have been there from day one.

Before this, the process was clunky. Users had to navigate a separate opt-out form. Now, the control lives right in the settings and can be flipped at any time. There’s a crucial detail, though: disabling the history doesn’t mean your data instantly vanishes. OpenAI will hold onto new conversations for 30 days, peeking at them only if it needs to sniff out abuse. After that, the company says, the data is permanently deleted. That 30-day window is a practical compromise, but it’s one users should absolutely know about before they start sharing anything truly confidential under the assumption of immediate privacy.

This move isn’t happening in a vacuum. Italy’s Garante briefly banned the service over data concerns, and regulators across the EU have been circling. A setting that puts data control a single click away is a much cleaner narrative for OpenAI than a labyrinthine help center article. And they’re not stopping with consumers. OpenAI also teased a ChatGPT Business tier, slated for the coming months, which will default to the API’s data usage policies — meaning end-user data won’t be used for model training. That’s the enterprise peace offering that companies waiting on the sidelines have been demanding.

Alongside the privacy tweaks, users are also getting a new export tool that packages up your ChatGPT data and sends it via email. It’s a nod to data portability that further rounds out the user-facing controls. Taken together, the changes address a straightforward problem: people want to use powerful AI without feeling like every prompt is a permanent donation to the machine. Whether this quiets the privacy critics for good is an open question, but it finally puts the control where it belongs — right in the user’s hands.