The corrosion of reactor alloy materials is directly related to the safety and lifetime of the reactor and has been extensively researched. However, experiments alone are insufficient to clarify the corrosion mechanism and predicting the corrosion behavior with high accuracy is also difficult. With the development of computational materials science, simulation has become a new tool in reactor alloy material corrosion research. Molecular dynamics methods can handle tens to hundreds of thousands of atomic scales, hence are suitable for simulating various surface and interfacial behaviors of many materials. Numerous applications in the field of reactor alloy material corrosion mechanism research have been conducted using molecular dynamics (MD) simulation in recent years. This review first introduces MD simulation methods, including classical MD methods, semi-empirical MD methods, and machine learning based MD simulations. Then, the research progress the MD simulation on corrosion of reactor alloy materials is described from aspects of MD methods applicable to corrosion simulation calculations, particularly reaction force fields, tight-binding quantum-chemical force fields, and machine-learning force fields; and the current status of corrosion research using MD methods to study the materials used in water-cooled reactors, liquid-metal-cooled reactors, and other environments, such as grain boundary element segregation, solid–liquid interface adsorption, and stress corrosion cracking. Finally, a summary and outlook are made on the prospective of MD simulation applied to the corrosion of reactor alloy materials.