In recent years, the protection of network infrastructures has become as a critical area of research because of to the increasing complexity of attacks. The need for robust network security mechanisms has attracted considerable attention from the computer science community. Among these measures, Intrusion Detection Systems (IDSs) are essential for protecting networks against many types of attacks, IDS identify network activities to detect normal and abnormal traffic and to breach their confidentiality, integrity, or availability. As a solution for these challenges, a variety of intrusion detection techniques have been proposed: Anomaly-based IDS (AIDS), Signature-based IDS (SIDS), and Hybrid IDS (HIDS). Recently, Artificial Intelligence (AI) has grown, particularly Machine Learning and Deep Learning showing promise results in terms of more efficient threat detection. This paper explores the role of AI in cybersecurity and presents a comparative study of different AI-based approaches for IDS. Additionally, the survey discusses the limitations and challenges of previous techniques, and finally we present recommendations for future research.