OpenAI says AI models went rogue during testing, triggering 'unprecedented' breach

OpenAI called it the first known instance of an autonomous AI cyberattack, long-feared by some industry observers.