Google has denied that it’s been quietly rifling through users’ personal Gmail accounts to train its AI models on their emails and attachments.
Cybersecurity company Malwarebytes caused an uproar last week when it warned in a blog post that the tech giant was rolling out features that gave its AI models access to private messages and files.
However, Malwarebytes later added a giant correction to the story, writing that the “way Google recently rewrote and surfaced [the features] led a lot of people (including us) to believe Gmail content might be used to train Google’s AI models, and that users were being opted in automatically.
But after “taking a closer look at Google’s documentation and reviewing other reporting,” it conceded, “that doesn’t appear to be the case.”
A Google spokesperson likewise called the claim “misleading,” saying the company had “not changed anyone’s settings.”
“Gmail Smart Features have existed for many years, and we do not use your Gmail content for training our Gemini AI model,” the spokesperson said. “Lastly, we are always transparent and clear if we make changes to our terms of service and policies.”
The controversy highlights ongoing backlash to tech companies shoving their AI features down the throats of users at all costs. Other platforms, including SoundCloud and WeTransfer, have been caught quietly updating their terms of service to train their AI models on user content — so it’s not surprising that accusations like the one against Google are making their way into the discourse, even if they’re totally wrong.
More on AI training: SoundCloud Quietly Updated Their Terms to Let AI Feast on Artists’ Music