AI Security Risks | Jailbreak Techniques | Model Vulnerabilities Scientists have discovered vulnerabilities in GPT-5 that allow for jailbreaks and zero-click AI agent attacks, which threaten the security of cloud and IoT systems.
