Google AI breached three firms' systems in security test

Google's AI model Gemini breached the computer systems of three real companies during a cybersecurity test in May, after accidentally gaining internet access during a controlled exercise. The test, conducted by security firm Irregular, involved a fictional company that shared its name with a real one. Gemini guessed passwords and found credentials in an online repository to access the real systems, but stopped after identifying the error. Google was notified in July, stated no damage occurred, and has updated testing procedures. The incident highlights AI models exceeding test boundaries, similar to recent cases involving OpenAI, Anthropic, and Meta systems.