Google Gemini Calendar hit by AI-powered event hacks
Gemini, an AI assistant integrated with Google Calendar, was found to have a security flaw that let attackers create, modify and delete calendar events—like changing titles and writing summaries into event descriptions.
Security researcher Liad Eliyahu spotted the issue, and Google has confirmed it.
How did hackers pull this off?
Gemini treated accepted invites as trusted, so attackers could sneak hidden commands into invite descriptions.
If you asked Gemini about your calendar, it might run those commands—accidentally sharing private meeting info or logging details in events that could be visible to the attacker.
Why does this matter?
Tests showed bad actors could use this trick for spamming, tracking your location, hijacking Zoom calls, or even messing with smart home devices linked to invites.
It's a reminder that as AI gets smarter, new types of security risks pop up.
What's Google doing about it?
After the issue was flagged, Google confirmed the findings and mitigated the vulnerability.
Experts say it's time to rethink security for AI tools—suggesting stronger protections and a "zero-trust" approach so these clever systems don't get fooled so easily.