Researchers Uncover GPT-5 Jailbreak and Zero-Click AI Agent Attacks Exposing Cloud and IoT Systems

Cybersecurity researchers have developed a novel jailbreak method that bypasses OpenAIs ethical guardrails in GPT-5, enabling the model to generate illicit instructions. By combining the Echo Chamber technique with narrative-driven steering, attackers can manipulate GPT-5 to produce undesirable outputs, raising concerns about the robustness of current safety measures in advanced large language models.