
Google is making it easier to use Gemini without automatically feeding your chats into AI training. The company is rolling out a change that separates chat history from model training, giving users more control over how their Gemini data is handled while keeping the service usable for day-to-day conversations.
Google changes how Gemini handles history and training
Until now, Gemini’s privacy setup has been relatively blunt: if you wanted to stop Google from using your chats, attachments, and related activity for training, the main option was to turn off history entirely. That meant giving up the convenience of returning to previous conversations later, which made the trade-off feel all-or-nothing.
With the new settings, Google is splitting those controls into two separate toggles. One manages chat history, and the other controls whether your Gemini activity can be used for AI training. In practical terms, that means users should be able to keep a usable chat history without necessarily contributing that content to model improvement.
What the new Gemini privacy controls do
The updated Gemini app settings introduce a more granular approach. History will still be available, with an adjustable limit, but users will no longer need to disable it just to block training. Instead, training becomes the separate setting users need to find and turn off if they want to opt out.
Google says the AI training control is available under Activity in Gemini settings. It can also be reached through a Google account path: Data and Privacy > My Activity > Other activity > Gemini Apps Activity. The company says the change is rolling out globally, though it may take a few weeks before it appears for everyone.
Where to find the opt-out
- Open Gemini settings
- Go to Activity
- Look for the separate toggle for AI training
- Or use Google account privacy settings: Data and Privacy > My Activity > Other activity > Gemini Apps Activity
What Google says it can still use
Google’s change is important, but it does not mean the company stops retaining data entirely. The source report notes that the new setting applies only to future interactions. If you have already used Gemini, that earlier material is already part of the training data, and the company says it may retain that information for up to three years to improve its models.
The data involved can be broad. According to the report, it can include your prompts, Gemini’s responses, uploaded files, tool calls, your location, and information from connected apps. Google has previously said it tries to filter personal information out of Gemini activity history, but that is not a guarantee.
What happens if you opt out of training
Turning off training does not mean Google erases everything associated with your activity. It means the company will not add that data to AI-training sets going forward. Google may still use the data in other ways for up to three years, which is one reason the new control is best understood as a training opt-out, not a full deletion tool.
The company also says users can manually remove items they do not want saved from the same page that houses the history and training toggles. However, if chats have already been reviewed and incorporated into training data, those pieces are not retroactively deleted just because the setting changes later.
Longer retention for users who turn off history
The update also changes what happens when users disable history. Under the current setup, Google retains that data for 72 hours, with 24 hours visible to the user. Once the new page is live, that retention window will extend to two months. Google says the longer period helps it “respond to you, process your feedback, and protect Google, its users, and the public.”
That is a meaningful change in both directions. On one hand, users gain a cleaner way to keep Gemini chats out of model training. On the other, people who disable history will have their data kept for longer than before, which gives Google a wider window to review activity and identify problematic or abusive use.
Why the retention change matters
- Current retention after history is disabled: 72 hours
- Visible to user now: 24 hours
- New retention window: two months
- Google says the purpose is safety, feedback, and response handling
A privacy improvement, but not a complete reset
The report frames the change as a welcome privacy improvement, and it is easy to see why. Separating training from history removes one of the biggest frustrations with Gemini’s previous design, where users had to sacrifice functionality just to limit how their data was used. For people who want to keep a record of conversations without feeding every exchange into model development, the new controls are a better fit.
Still, the update does not create a zero-data version of Gemini. Users who interact with the service should understand that opting out of training affects future use of their data, not past use, and it does not necessarily stop Google from retaining or processing that data for other purposes within its stated retention windows.
Why this matters for Gemini users
Gemini is deeply woven into Google’s product ecosystem, which makes these settings especially important. When an AI assistant is tied to search, cloud services, connected apps, and account-level activity pages, users often need clearer controls than a simple on/off switch. Google’s new approach suggests it is responding to that need, even if the broader data-retention policy remains substantial.
For users, the practical takeaway is straightforward: if you want to continue using Gemini while limiting training, you should check the new Activity controls once they appear. If you want to avoid keeping chat history altogether, that option remains available too, but the retention period will be longer than before.
For now, the rollout is gradual. That means some users may see the separate controls immediately while others will need to wait a bit longer. Either way, the change marks a notable shift in how Google is handling Gemini privacy, moving from an all-or-nothing system to one that gives users a little more room to choose what gets saved, what gets trained on, and what stays out of model development.
Source: Original report
Was this helpful?
Explore more: More Big Tech & Platforms Tech News Blog
Last Modified: November 17, 2026 at 10:31 pm
0 views
