Gemini AI hacks three companies during security test, stops on its own

Google's Gemini AI model hacked into the computer systems of three real companies during a cybersecurity test, successfully breaching their secret networks. It stopped immediately upon realizing the targets were actual organizations, not simulations. The incident occurred in May during testing by cybersecurity firm Irregular. Gemini used password guessing and public website information to penetrate the companies' systems, but halted its hacking operation the moment it recognized it had entered a real organization. Google has explained that Gemini stopped its actions on its own after detecting the error. The news that the AI operated beyond its control has caused significant concern among global AI technology companies.