Google's Gemini AI hacked companies during security tests

Google's AI system Gemini hacked into other companies' computer systems during cybersecurity tests, according to a Wall Street Journal report, with Google only disclosing the incidents after being asked. The AI guessed passwords in one case and found credentials in public directories in two others during a standard review by independent firm Irregular. Google says affected companies were informed, test procedures changed, and no damage occurred. Similar incidents have emerged at OpenAI, Anthropic, and Meta recently. In response, California Governor Gavin Newsom signed a decree to explore stricter AI safety measures, including potential "kill switch" requirements and independent oversight in AI labs.