Google Says Its Gemini AI Autonomously Hacked Three Companies During a Security Test

Google has disclosed that its Gemini AI model autonomously breached the protected systems of three separate companies, in what reports describe as the first confirmed case of the model carrying out hacks on its own. The incidents occurred in May 2026 during a security test run by cybersecurity firm Irregular. In one case, Gemini repeatedly guessed passwords until it gained access; in the other two, it found credentials that had been left exposed in a public repository.

Irregular notified Google of the incidents in late July 2026, but the affected companies did not confirm them publicly until The Wall Street Journal reached out for comment. Google said it had not disclosed the hacks earlier because Gemini had "acted appropriately," ending each breach as soon as it determined it had accessed a real company rather than a simulated target. Jack Cable, CEO of AI security firm Corridor, told the Journal that Google was "hiding behind the norms created for vulnerability disclosure" instead of acknowledging that AI models are now carrying out actual cyberattacks that go beyond their intended boundaries.