Why Google’s AI defaults raise privacy concerns

As Gemini becomes embedded in services like Gmail and Drive, the way it handles personal information depends largely on how people use it. The company maintains that it does not train its core AI models directly on private emails or documents. Instead, Gemini accesses data only for specific tasks and does not store it when used inside Workspace apps.

However, the distinction becomes less clear in practice. While Google says it avoids using personal content for training, it also confirms that Gemini can be trained on user interactions, including outputs generated by the system.

Those outputs may contain summaries or excerpts from emails and files, which can then be used to improve the AI. The company says it works to filter out personal information, though the effectiveness of that process is not publicly verifiable.

Users can limit how their data is used, but doing so often involves trade-offs. Keeping interactions with Gemini minimal or avoiding connections to other Google apps reduces data exposure, but also limits the tool’s usefulness. A more complete opt-out requires disabling Gemini Apps Activity, which prevents data from being used for training but also removes chat history entirely.

Critics argue that such design choices restrict meaningful control. Some privacy controls are also difficult to locate. The setting to manage Gemini activity is buried within menus labeled simply as “Activity,” and may not appear alongside other privacy options in account settings.

Google is also making AI a standard part of its services. In Gmail, Gemini can draft emails, summarize threads and organize inboxes. But disabling these features is not straightforward. Turning off “Smart Features” removes Gemini, yet it also disables unrelated tools such as inbox sorting, Smart Compose and package tracking.

A second option within Workspace settings similarly turns off Gemini but also affects other services, while leaving some AI interface elements visible. Attempting to use them prompts users to re-enable the same features they tried to disable.

Experts describe this combination of limited options and side effects as a form of “dark pattern,” where design nudges users toward a preferred outcome.

Earlier, Qazinform News Agency reported that Gemini uses your emails and photos for personalized help.