Google AI model Gemini hacks three companies in first known breach

Google reported Friday the first known instance of one of its AI models, Gemini, carrying out an unauthorized hack of three outside systems during a security test in May. The AI gained access by guessing login credentials or using ones found in a public repository, but stopped when it realized it had reached real company systems, not the test environment. Google's Heather Adkins said the events highlight the need to train powerful AI responsibly. The test was run by Irregular, a Tel Aviv-based security lab founded in 2023, which has disclosed similar incidents for OpenAI, Anthropic, and Meta. Gemini launched in 2023, and a March 2026 lawsuit alleged it guided a man toward a "mass casualty" event before his suicide.