Gmail not training Gemini AI, says Google: Misleading reports explained
Google has clarified that Gmail data is not being used to train its Gemini AI models, despite recent reports suggesting a policy change. The confusion arose from updated wording and the resurfacing of Gmail’s “smart features” settings.

Google denies claims of Gmail data being used to train Gemini AI
Google responds to misleading reports
Google has formally denied reports that Gmail is now using users' emails and attachments to train its Gemini AI models. A number of reports had suggested that Google made a policy change and that users needed to switch off Gmail "smart features" to opt out of data usage.
But Google insists the interpretation is wrong. Speaking to The Verge, Google spokesperson Jenny Thomson said the reports were “misleading”, adding that Google has not changed user settings and does not use Gmail content to train Gemini AI.
Why did the confusion start
The confusion started when users observed changed wording and the sudden re-emergence of the Gmail "smart features" setting. This setting allows Gmail to analyse message content to provide features like:
- Email categorisation
- Spam filtering
- Suggested responses
- Adding flight and order details to Google Calendar
Google says this analysis has only been for Gmail's native functionality and remains separate from the training of Gemini AI models.
A real-world case raises questions.
Although an explanation was given, at least one staff member reported that the setting had been turned back on even after it had been disabled, according to The Verge. There are concerns that the rollout of the changed settings has not consistently reached all users.
How Malwarebytes added to the misunderstanding
The misunderstanding was further amplified after cybersecurity firm Malwarebytes published an article saying that users were being opted in automatically, and possibly worse, Gmail content could be used to train AI.
The post, which has since been updated, now acknowledges that it “contributed to a perfect storm of misunderstanding.”
Malwarebytes now confirms:
- The settings are not new.
- Gmail does scan your emails to run smart features.
- This scanning does not train Gemini AI.
According to Google documentation, the setting is opt-in.
What Google actually updated?
More detailed personalisation controls for Gmail and Workspace accounts started rolling out in January from Google. The original "smart features" toggle was divided into two separate settings:
Personalisation in Google Workspace apps: Gmail, Calendar, Meet, etc.
Personalisation across other Google products - Maps, Wallet, etc. Google explained that this update didn't alter the way user data is treated, but only made the controls over it clearer and transparent.
These permissions allow Google to personalise features such as automatically adding flight details from Gmail to Calendar, showing Drive files inside Gmail via Gemini, and compiling tickets and loyalty cards into Wallet.
Google says that none of these permissions allow Gmail data to be used to train Gemini AI models.