Google has responded to viral claims that it trains its AI based on users’ Gmail data. In particular, a recent report by Malwarebytes claimed that Google was changing settings to automatically train its AI on users’ emails. The report also claimed that users must turn off smart features and personalization settings to prevent Gmail from training its AI on emails.
The report sparked a barrage of angry comments on social media, with users outraged that Google would train an AI with their personal data without their explicit permission. And it turns out Google doesn’t do that.
In a social media post, the company said it has not changed any settings and does not use Gmail content to train Gemini AI.
“Let’s set the record straight about recent misleading reports. Here are the facts,” Google said in a post to X.
“We are not changing anyone’s settings. Gmail smart features have been around for years. We do not use your Gmail content to train Gemini AI models. We are always transparent and clear when we make changes to our terms of service and policies,” the tech giant added.
Malwarebytes has since updated its report to confirm that Google does not, in fact, train its users on email, and blames Google’s ambiguous language for the misunderstanding.
“The setting itself is not new, but the way Google recently rewrote it and made it public has led many (including us) to believe that Gmail content is being used to train Google’s AI models and that users are automatically opted in. After taking a closer look at Google’s documentation and considering other reports, that doesn’t appear to be the case,” Malwarebytes said in its latest report.
Notably, this isn’t the first time Google has had to deny false reports about Gmail this year. Google issued the clarification in September after multiple media outlets reported that 2.5 billion Gmail users were compromised in a breach. The company said these reports are false and causing unnecessary panic.
Where Google trains data
While Google denies training Gemini on users’ Gmail content, the company does automatically train Gemini on the conversations users have with its AI chatbot. This means that to prevent Google from training AI models based on Gemini conversations, users must turn off the activity settings in the Gemini app in their Google account.
However, Google isn’t alone in this effort. Anthropic recently informed users that it will train conversations with Claude by default. Meanwhile, Meta also announced that it will display ads and personalized content based on the conversations users have with its chatbot.
