Artificial Intelligence (AI) and Machine Learning (ML) are revolutionizing the field of nuclear sciences by introducing advanced methods for reactor physics, nuclear data analysis, experimental design, and operational monitoring. Recent breakthroughs span the application of deep learning, surrogate models, and physics-informed neural networks (PINNs) to critical challenges such as neutron transport simulations, reactor core design optimization, real-time anomaly detection, uncertainty quantification, and predictive modeling of transients and fuel burnup. These data-driven approaches enable improved accuracy, efficiency, and real-time decision support while addressing traditional computational bottlenecks. Despite significant progress—especially in leveraging PINNs to embed physical laws within machine learning frameworks—challenges remain in model generalization, interpretability, industrialization, and comprehensive validation. This review synthesizes current AI applications in nuclear sciences, identifies key advances, and highlights challenges and opportunities for future research.