Google denies claims that it uses Gmail emails to train Gemini AI. [Photo: X]
Google has responded to viral rumours claiming that it is training its Gemini AI using Gmail users’ emails. The company issued a firm clarification after social media posts and a widely circulated report suggested that Gmail had quietly changed its settings to allow AI training. Google denied the claims and said it had made no changes to user settings or data policies.
Google has issued a strong clarification after viral online reports claimed the company was training its Gemini AI model using users’ Gmail data. The tech giant denied the allegations and said it had not changed any user settings.
A recent report by Malwarebytes had suggested that Google was automatically using Gmail emails and attachments to train its AI. The post also claimed that users had to disable Smart features and personalisation to stop Gmail from feeding data to AI systems. The claim spread quickly on social media, leading to outrage from users who feared that Google was scanning their private emails for AI development.
However, Google has now confirmed that none of these claims is true. In a post on X, the company wrote: “Let's set the record straight on recent misleading reports. Here are the facts.”
Google further 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.”
Malwarebytes later updated its report and accepted that the confusion came from Google’s unclear wording. It said the settings were not new, but their recent placement made it appear as if Gmail content could be used to train AI.
The updated report noted:
The misunderstanding grew after some users saw their smart features toggle reappear, although Google has not confirmed any reset.
Earlier this year, Google introduced more detailed controls for Gmail and Workspace users. It split the old smart features toggle into two options:
According to Google, none of this data is used for training Gemini.
While Gmail data does not train Gemini, the company has confirmed that conversations with the Gemini chatbot can be used for AI training unless users disable the Gemini Apps Activity setting. Google is not the only company doing this.
Anthropic recently announced that it trains Claude AI using user chats by default. Meta has also said it will show personalised content and ads based on conversations with its chatbot.
This is the second major Gmail-related rumour Google has had to deny this year. In September, the company dismissed reports claiming that 2.5 billion Gmail accounts were leaked. Experts say such incidents show a growing trust gap between tech companies and users.
As AI grows, users want clarity about how their data is used. Any ambiguity, even in settings, can quickly turn into panic — as seen with the Gemini rumour.
