Google denies viral claims it is using your emails to train Gemini AI

Gmail

Google is disputing viral social media posts and reports suggesting the company has quietly updated its policies to allow Gmail messages and attachments to train its Gemini AI models. A widely shared article from Malwarebytes fuelled concerns by claiming users could only opt out by disabling Gmail’s “Smart Features”, including spell check.

But Google spokesperson Jenny Thomson told The Verge these claims are inaccurate. She says the company hasn’t changed any user settings and does not use Gmail content to train Gemini. Smart Features, she adds, have been part of Gmail for years, powering conveniences like package tracking and calendar auto-adds. "We do not use your Gmail content for training our Gemini AI model," she told The Verge.

Users still should recheck settings

Despite Google’s denial, some confusion remains. One staff member at The Verge noted they had previously disabled some Smart Features, only to find them enabled again. Google updated its personalisation settings in January, allowing users to manage smart features separately across Workspace and other Google products such as Maps and Wallet. That update may have contributed to the renewed uncertainty.

When enabled within Workspace, Smart Features allow Google to use your Workspace activity to personalise your experience. Google confirms this does not equate to using email content for training any AI model, including Gemini.

What are smart features?

Turning on these features unlocks conveniences like suggested actions, order tracking, and quick-add travel reminders from Gmail to your calendar. Turning them off limits personalisation but stops Gmail from processing content for these specific enhancements. Google stresses that enabling them does not give the company permission to feed email text or attachments into model training pipelines.