Beware the Invisible Trap: How Google Gemini’s AI Email Summaries Could Be Hijacked by Phishers
Google’s push to inject its Gemini AI into everything from Search to Chrome is full throttle—yet not without some hiccups. One Gemini Workspace feature that summarizes emails might be easily fooled by a sneaky phishing trick, according to a bug report spotted by BleepingComputer.
Here’s how the scam works: an attacker hides malicious instructions in plain sight—white text on a white background at the end of an email—making it invisible to human eyes but fully readable by Gemini. The AI then dutifully includes a fake warning in its summary, telling you your password’s been compromised and urging you to call a shady phone number. The “Admin” tag buried in the invisible text forces Gemini to copy the scam message verbatim.
Thankfully, this is only a theoretical flaw for now and only affects Workspace users, not everyday Gmail accounts. But the ease with which AI can be tricked into spreading phishing content is a big red flag. As more people rely on AI to read and act on their emails, these kinds of exploits could become more common.
So, before you trust any AI-generated alerts about your account security, take a moment to verify. After all, when AI meets hackers, it’s the user who needs to stay sharp.