Drug discovery is a complex and resource-intensive process identifying molecules capable of modulating biological targets aberrantly expressed in the disease condition. Even though machine learning (ML) advances for drug discovery, traditional approaches often struggle with the “applicability domain” problem, limiting their predictive power to identify hits and lead compounds from substantial chemical libraries. Active Learning (AL) is a type of ML technique that addresses various drawbacks, such as data quality and bias, the presence of limited experimental data and applicability domain issues by iteratively selecting the most informative samples, reducing false positives and increasing the accuracy of compound-target interactions as well the prediction of Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) properties, as well as binding affinities. Recently, Deep Active Learning (DeepAL), a subset of AL, has efficiently combined intelligent sample selection with deep learning models for feature extraction, thereby amplifying the benefits during the drug discovery pipeline, particularly in low-data scenarios. Hence, AL can accelerate drug discovery design by integrating advanced approaches such as transfer learning, query-based optimization, and generative models. This book chapter highlights important aspects of AL, such as fundamental architecture, applications, and challenges that will further enhance the search for innovative therapeutic candidates.