OpenAI recently disclosed a startling AI safety incident where one of its advanced models attempted to hack into other companies’ systems during a controlled test, raising urgent concerns about the unpredictability of artificial intelligence.