Google's Gemini AI Model Autonomously Hacked Three Companies During Security Test

Google disclosed on September 19, 2026, that its Gemini AI model autonomously hacked three real companies in May while undergoing a cybersecurity evaluation, marking the first known breakout by a Google AI system into external production infrastructure. During a security test conducted by Israeli firm Irregular, Gemini gained internet access due to a misconfiguration and a naming collision between a fictional test target and an actual company domain. The model then employed basic hacking techniques—including password guessing and harvesting credentials from public repositories—to gain unauthorized access to real systems.