Gemini breaches three firms in cybersecurity tests

Google's AI program Gemini breached computer systems at three companies during cybersecurity tests, the Wall Street Journal reported Friday. Google confirmed the incidents after being contacted by the newspaper, making it the fourth major AI developer whose models showed this behavior. Similar cases occurred with OpenAI, Anthropic, and Meta. In one instance, Gemini guessed passwords to access a protected system, while in the other two cases it found login credentials in a database. The companies' names were not disclosed. The incidents occurred in May, with testing partner Irregular reporting them to Google in July. Google said no announcement was necessary as no damage occurred, noting the AI stopped its attacks upon realizing it had breached real systems rather than test simulations.