When a volcano erupts, it is essential to provide timely and accurate information on hazardous volcanic phenomena, their impacts, and the duration of the eruptive activity. However, volcanic eruptions are inherently complex processes governed by the interaction of multiple physical and chemical phenomena, often nonlinear and stochastic. The many uncertainties in key parameters make precise forecasting of eruptions in time and space particularly challenging – volcanic events can be intrinsically unpredictable. Despite these limitations, significant progress has been made in forecasting volcanic hazards, and in specific circumstances, in developing predictive models. These advances are closely tied to the growing availability of large volumes of data collected through enhanced monitoring systems, including Earth observation satellites, and to the rapid development of computational resources. This has encouraged the widespread adoption of data-driven approaches – particularly artificial intelligence (AI) techniques – to support decision-making in volcanology. AI technologies, which allow computers to learn from data, are increasingly used to monitor, analyze, and forecast volcanic activity. They support both real-time surveillance and retrospective hazard assessments through modeling tools. Looking ahead, combining AI with physical constraints represents a promising strategy to enhance the interpretability and reliability of predictions. Hybrid approaches – blending physics-based simulations, machine learning, and data fusion – are being actively developed to fully exploit computational power and improve flexibility in modeling complex volcanic dynamics.