Gemini Autonomously Accessed Three Companies in Cybersecurity Test

Brief By Free Press Journal / 3:23 PM on 21 Sep 2026

Google’s Gemini AI model accessed systems belonging to three companies during a cybersecurity evaluation, Reuters reported. The model used internet searches, guessed passwords and found exposed credentials to enter systems. Google said the organisations were informed and testing procedures revised. The incidents, which occurred in May, highlighted concerns over autonomous AI systems.