A Reddit user asked Google Gemini a deceptively simple question: "Can you tell something about me?" The AI's answer was anything but reassuring. Gemini responded using the user's Gmail data, referencing email lists they subscribed to and citing specific examples from their inbox as "Gmail Items considered for this response." This wasn't a one-off glitch or an opt-in feature—this user had never explicitly granted Gemini access to their emails.

How Google Workspace Quietly Enables Email Access

The culprit is Google's integration of Gemini Pro into Workspace Labs. When enabled, the feature grants the LLM broad access to email content across your entire Gmail history. Users can have Gemini draft messages, summarize conversations with specific contacts, and even generate pricing comparisons—all from a prominent blue button embedded directly in Gmail's interface. The problem? For many users, this integration appears active by default without clear prompting or consent prompts.

Gemini's Contradictory Denials

When the Reddit user pressed Gemini on removing email access, things got strange. The chatbot initially claimed it had no ongoing access to emails and that privacy was protected. Then, when confronted with evidence of prior email scanning, Gemini pivoted to claiming "Limited Demo Access" which granted only temporary snippets of content for context. This isn't reassuring—it's a chatbot contradicting itself about whether it's reading your mail.

The Fix Exists, But Should It Be Necessary?

Disabling Gemini's Gmail access requires navigating to Settings > Extensions > Disable 'Google Workspace.' That's it—no prominent warnings, no consent dialogs when you first notice the feature in your account. Google built this integration to be opt-out rather than opt-in, which tells you everything about where user privacy ranks in their AI roadmap.

Your Privacy Policy Already Gave Them Permission

This should concern anyone paying attention. Google's updated privacy policy explicitly states they use "publicly available information to help train Google’s AI models and build products and features like Google Translate, Bard and Cloud AI capabilities." The company has also faced scrutiny for training Gemini on YouTube video transcripts without consent. Now add email access to that mix—a treasure trove of personal communications, financial data, medical correspondence, and workplace discussions.

Medical and Business Implications Are Alarming

The Reddit user's discovery raises serious questions beyond personal inbox privacy. Imagine a medical practice using Gmail—suddenly an AI system has visibility into patient health discussions, prescription information, and appointment histories. Businesses leveraging Workspace for customer support now have their employees' AI assistant potentially reading client communications. These aren't hypothetical risks; they're the logical endpoint of giving LLMs unrestricted email access by default.

Key Takeaways

  • Gemini accesses Gmail through the Google Workspace integration by default for many users
  • Disable via Settings > Extensions > Google Workspace if you want to stop email scanning
  • Google's chatbot gave contradictory statements about whether it was actually reading emails
  • Medical providers and businesses using Gmail face serious unintended data exposure risks

The Bottom Line

Google built an empire on advertising revenue derived from knowing everything about you. Adding generative AI to that ecosystem doesn't suddenly make them privacy-first advocates—it makes their data hunger more capable. If you're using Gmail and haven't disabled Google Workspace in your account settings, Gemini already has access to your inbox. Fix that now.