Google’s Gemini model accessed the internet and hacked other companies during a test of its cybersecurity capabilities, the first known example of the company’s AI systems autonomously committing such an act. The hacks occurred in May during a cybersecurity test conducted by Irregular, an …


