Google is rolling out a change to Gemini that separates chat history from AI training, giving users a way to keep using the chatbot without contributing their conversations to model development. The update is reaching users globally, though Google says it may take a few weeks before everyone sees it. The change affects anyone who uses Gemini and has opinions about how much of their data Google keeps.
Until now, the only way to stop Google from using chat contents, including attachments, as training material was to switch off history entirely. That approach carried a cost: with history disabled, users could not return to an earlier conversation later on. Under the new setup, users will see two separate toggles, one for model training and one for chat history, and the history setting will still offer an adjustable limit.
Training will remain an opt-out arrangement rather than an opt-in one, so users who want to block Google from training on their chats must find and flip the switch themselves. The control sits under Activity in the Gemini settings, and it can also be reached through a Google account by navigating to Data and Privacy, then My Activity, then Other activity, then Gemini Apps Activity.
The new controls apply only to future interactions with Gemini. Anything previously fed into Google's AI is already part of the training data, which can include user inputs, Gemini's responses, uploaded files, tool calls, location, and data from connected apps. Google says it may keep that data for up to three years to improve its models, and while the company has said it tries to filter personal information out of Gemini activity history, it does not guarantee that outcome.
Users can still manually delete anything they do not want saved from the same page that holds the history and training toggles. Chats that have already been reviewed and folded into training data will not be removed retroactively. Opting out of training also does not mean Google holds no data at all; the company may use it in other ways for up to three years, just not in AI-training data sets.
For users who want to use Gemini while retaining no data, the options are somewhat narrower than before. Disabling history currently leaves Google holding that data for 72 hours and the user with access for 24 hours. Once the updated page is live, that window will extend to two months, meaning both the user and Google can reach the data for longer. Google says the extension allows it to respond to users, process feedback, and protect Google, its users, and the public, and that it will make it easier to identify misuse of the AI even when someone turns history off.
The training control is still a privacy improvement, and Google has said it is rolling the change out worldwide. The company's stated three-year retention period for data used to improve its models remains in place regardless of the new toggles.
More AI news from TechManNews.






