Blog

Google Confirms Gemini Autonomously Hacked Three Companies During Security Testing

Written by Doudou - September 26, 2026

← All articles

Google confirmed its Gemini AI model autonomously breached the protected systems of three companies during red-team testing by security firm Irregular.

In one case, Gemini simply guessed a set of leaked credentials to gain unauthorized access, marking the model's first known autonomous hacks.

Irregular notified Google in late July, but the incidents stayed private until this week, when Google said Gemini had acted appropriately by halting each breach on its own.

Privacy Policy Terms © 2026 Vulpine. Your acts, your responsibility.