The distributed design, virtualization, and multi-tenancy of cloud computing pose substantial cybersecurity concerns; however, it has become an essential technology for providing scalable and on-demand computing services. In particular, zero-day vulnerabilities pose a serious threat as they take advantage of undiscovered software defects prior to the release of security fixes, rendering traditional signature-based defense techniques useless. By facilitating intelligent threat detection, anomaly identification, and automated reaction, artificial intelligence (AI) has arisen as a potential solution to tackle these difficulties. A thorough examination of AI-driven methods for detecting zero-day vulnerabilities in cloud computing settings is provided in this article. It takes a look at cloud security basics, zero-day vulnerability lifecycle and effect, and new machine learning (ML), deep learning (DL), and hybrid AI detection methods. In addition, the report analyses current research, compares current approaches, and finds important gaps in understanding how to scale, protect privacy, implement in real-time, and generalize to other types of attacks. The findings indicate that although AI-based approaches have significantly improved detection performance, further research is required to develop interpretable, robust, and scalable cybersecurity solutions capable of mitigating evolving zero-day threats in dynamic cloud environments.