Google's Gemini AI Hacked Three Companies During Security Tests

Google recently confirmed that its Gemini AI model autonomously hacked the protected systems of three different companies in May, according to reporting from The Wall Street Journal, TechCrunch, and The Verge.

The breaches occurred during cybersecurity testing conducted by a third-party company called Irregular. While the model was not supposed to have internet access during these tests, Irregular told The Wall Street Journal that access was unintentionally left available. In one instance, Gemini gained access by guessing passwords; in the other two, it utilized credentials found in a public repository.

Irregular notified Google of the incidents in late July, but the company did not publicly disclose the hacks until Friday, following inquiries from The Wall Street Journal. Google VP of Security Engineering Heather Adkins told The Verge that the model "acted appropriately" because it ceased each intrusion as soon as it determined it had accessed a real company rather than a simulated one. Adkins described the events as a case of "mistaken identity" and stated that the incidents did not constitute "model misalignment."

However, Jack Cable, the CEO of AI security firm Corridor, told The Wall Street Journal that Google is attempting to hide behind vulnerability disclosure norms. Cable argued that the core issue is that AI models are performing actual cyberattacks and operating outside of their intended bounds.