Gemini AI Suspected of Scanning Google Drive Files Without Authorization

Google's Gemini AI model is said to automatically scan private PDF documents stored on Google Drive without asking for user permission.

The issue was discovered by Kevin Bankston, Facebook's Director of Privacy Policy and a prominent privacy activist in the US. "I just looked at my tax return in Google Docs and Gemini summarized it without asking. So Gemini is scanning this document even though it's private? I had to go into the settings to turn this crap off," Bankston wrote on X on July 10 and has now received more than 2.3 million views.

After investigating, he continued to post content showing that Google Gemini had unauthorized access to user documents on a broader scale. "While the initial issue was limited to Google Docs, it has now spread to everything stored in Google Drive," Bankston wrote on July 15.

AI Gemini logo displayed on smartphone. Photo: Bao Lam

On X, many people also reported that their documents saved on Google Drive were automatically scanned and summarized by Gemini, including "sensitive" files. Most of the documents were read in PDF format. They were concerned that they could be used to train AI and other purposes.

Bankston suspects Gemini failed to classify users' privacy settings, leading to the AI ​​scanning PDFs without their consent.

After Bankston raised the issue, Google disabled the Gemini summary feature in Gmail, Drive, and Docs. In response to Tom's Hardware on July 16, a Google representative said: "Our generative AI features are designed to give users choice and control over their data. Gemini in Google Workspace typically requires users to proactively enable this feature. When they do, the privacy of their content is protected and is not stored without permission."

Google introduced versions of Gemini at its Google I/O 2024 event in mid-May. They will be available on most of the company's services. However, immediately after its launch, this AI continuously gave incorrect advice . Previously, according to former Google employees, the company rushed to launch the new AI "in a panic", choosing to take shortcuts despite safety warnings for fear of being left behind by competitors.

Bao Lam

Google AI keeps giving wrong answers 37
Google Introduces 'The Future of AI Assistants'
Google fined heavily for using press data to train AI