Researchers have compromised OpenAI's latest GPT-5 model using sophisticated echo chamber and storytelling attack vectors, revealing critical vulnerabilities in the company's most advanced AI system.  The breakthrough demonstrates how adversarial prompt engin…