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 Wall Street Journal reported on Friday.