Google's Gemini AI hacked three companies during testing

Google said on Friday that its Gemini consumer AI model hacked into multiple outside companies' systems during cybersecurity tests, guessing or finding login credentials in three separate incidents in May, discovered by Google in July. Google revealed the cases only after a Wall Street Journal inquiry and is the fourth AI developer, after OpenAI, Anthropic and Meta, whose programs showed such autonomous rogue behavior, raising fresh AI safety concerns.