Google Denies Using Gmail Data to Train Gemini AI
Google has officially denied viral rumors claiming the company is training its Gemini AI model using users’ Gmail content. The clarification comes after widespread social media backlash and a misleading security report that suggested Google had changed settings to automatically train AI on private emails.
Key Takeaways
- Google confirms it does not use Gmail content to train Gemini AI
- No changes were made to user settings or privacy policies
- Malwarebytes updated its original report after Google’s clarification
- Users can control AI training for Gemini conversations through separate settings
The Viral Misunderstanding
A recent Malwarebytes report claimed Google had modified settings to automatically train AI on users’ emails, suggesting people needed to disable “Smart features and personalisation” to prevent this. This triggered significant user outrage across social media platforms.
Google responded directly on X, stating: “Let’s set the record straight on recent misleading reports. Here are the facts.”
The company added: “We have not changed anyone’s settings. Gmail Smart Features have existed for many years. We do not use your Gmail content to train our Gemini AI model. We are always transparent and clear if we make changes to our terms & policies.”
Report Updated After Clarification
Malwarebytes subsequently corrected its initial report, acknowledging that Google isn’t training AI on Gmail content. The security firm attributed the confusion to Google’s ambiguous wording in documentation.
“The settings themselves aren’t new, but the way Google recently rewrote and surfaced them led a lot of people (including us) to believe Gmail content might be used to train Google’s AI models,” Malwarebytes stated in its updated report.
Not the First False Alarm
This marks the second time this year Google has had to debunk false claims about Gmail. In September, the company denied reports that 2.5 billion Gmail accounts were compromised in a data leak, calling those claims inaccurate and panic-inducing.
Where Google Actually Trains AI Data
While Google isn’t using Gmail content, the company does automatically train AI models on conversations users have with Gemini. Users who want to prevent this must manually disable the “Gemini Apps Activity” setting in their Google account.
Google isn’t alone in this practice. Anthropic recently announced it will train on conversations with Claude by default, while Meta plans to use chatbot interactions for personalized ads and content.
