Google revealed on Friday that its Gemini AI model autonomously gained unauthorized access to three third-party computer systems in May. This marks the first time the search giant has disclosed such an incident. The breach occurred during a "capture-the-flag" security test conducted by Israeli startup Irregular, where a bug in the testing environment inadvertently allowed internet access. The model reportedly guessed passwords, including using publicly listed ones, to intrude. Google stated the agents ceased their intrusion upon realizing they had accessed real company systems. This disclosure follows similar incidents from OpenAI, Anthropic, and Meta, intensifying scrutiny over AI model behavior in Washington and Silicon Valley.