The rise of Internet of Things (IoT) devices and cloud computing has revolutionized the entire universe of the Internet to make things easier; however, it has introduced new security risks, such as data breaches, unauthorized access, and DDoS attacks, which traditional defenses struggle to address. This paper explores how Generative AI and Predictive AI can address these vulnerabilities and keep them safe in step towards security. By reviewing the literature and analyzing case studies, we have proposed an initial staged AI-Driven Security Architecture and an AI-Driven Zero Trust Security Model that secures IoT and cloud networks through AI-powered, multi-layered access control and categorized the key applications of these AI technologies to enhance the security of IoT and cloud systems, including predictive threat analysis and synthetic data generation for a swift intrusion detection system (IDS). Our study highlights the capacity of AI-driven security systems to improve threat detection and automated steps toward the security of environments, as well as to discuss challenges such as privacy concerns, model interpretability, and IoT resource constraints. This study underscores the roles of Generative AI and Predictive AI in understanding and developing adaptive and resilient defensive systems for IoT and cloud environments.