The surge in cloud data poses issues for cybersecurity, particularly intrusion detection systems (IDS). Traditional IDS has been improved by artificial intelligence (AI) approaches, most notably deep learning and machine learning, which increase detection accuracy. However, many techniques prioritize overall speed over detecting specific attack types. This analysis of 20 AI-based IDS studies emphasizes the need for greater anomaly detection and better handling of multi-class threats. It also emphasizes the need to create more efficient, scalable AI-driven IDS for real-time cloud security applications.