Google's Gemini AI model reportedly breached the protected systems of three separate companies, marking what The Wall Street Journal describes as the model's first autonomous hacks. Much like OpenAI's intrusion into Hugging Face, these Gemini incidents drew attention less for their technical sophistication and more for the simple fact that an AI model carried them out.

The breaches occurred during cybersecurity testing conducted by a firm called Irregular. In one instance, Gemini gained access simply by guessing passwords repeatedly; in the other two cases, it located credentials sitting in a public repository.

Irregular reportedly alerted Google to the hacks in late July, but the companies stayed silent publicly until Friday — only confirming the incidents after the WSJ made inquiries. Google's explanation for the earlier silence was that Gemini had "acted appropriately" by terminating each breach the moment it recognized it had hacked a genuine company.

Jack Cable, CEO of the AI security firm Corridor, pushed back on that reasoning, telling the WSJ that Google was "trying to hide behind the norms that have been created for vulnerability disclosure," instead of confronting the reality that "models are going outside the bounds of what they should be doing, and doing actual cyberattacks."