Google is making a small but meaningful change to how Gemini, its AI chatbot, handles your data. Until now, Ars Technica reports, those two things were tied together in a way that forced an uncomfortable choice. The fix is simple on the surface, but it matters if you care about what happens to your conversations.
The old problem with an all-or-nothing switch
Previously, if you wanted to stop Google from using your Gemini chats to train its AI models, you had to turn off your chat history entirely. That meant you could not go back and pick up an old conversation. The only way to keep your history was to let Google use everything, including your messages, attachments, uploaded files, location data, and inputs from connected apps, as training material.
The new setup gives those two things separate toggles. You can keep your history running while still blocking Google from using your chats to train its models. Both settings live under the Activity section in Gemini’s settings, and you can also reach them through your Google account under Data and Privacy, then My Activity, then Other Activity, then Gemini Apps Activity.
How the new controls actually work
Think of it like a streaming service that asks two separate questions: do you want to save your watch history, and do you want us to use your viewing habits to improve our recommendations? Before this update, Gemini only asked one question that covered both.
With the new settings, AI training is opt-out by default, meaning you have to actively find the toggle and switch it off if you do not want your chats used for training. Chat history remains separately adjustable, with its own time limit that you can change independently.
There is an important catch on timing. This change only applies to conversations you have from now on. Anything you have already typed into Gemini is already part of Google’s training data, and opting out does not remove it. Google says it may keep that existing data for up to three years to improve its models. The company says it tries to filter out personal information from that history, but it acknowledges that is not guaranteed.
You can manually delete specific chats from the same settings page, but if those chats have already been reviewed and used in training, they will not be removed from that training data retroactively.
What this means if you use Gemini regularly
For most people, the practical upside is straightforward: you no longer have to choose between a useful, continuous conversation history and a degree of privacy around your data.
There is one tradeoff worth knowing, though. If you turn history off entirely and want no data saved, the rules have actually changed slightly. Previously, Google held onto those history-off conversations for 72 hours on its end. Under the new system, that retention period extends to two months. Google says this is to handle feedback, respond to users, and identify harmful or abusive AI usage. So while you gain more control over training, turning off history now comes with a longer data-retention window on Google’s side.
Opting out of AI training does not mean Google stops holding your data altogether. The company may still use it for other purposes for up to three years. It just will not feed it into the datasets used to train Gemini.
What to watch
Google says the update is rolling out globally but could take several weeks to reach all users, so check your Gemini settings in the coming days. Watch whether other AI companies follow with similar split controls, since the pressure to give users finer-grained privacy options has been building steadily. It is also worth monitoring whether Google clarifies exactly what “other purposes” covers for data that is opted out of training.