The global pursuit of sustainable energy solutions has increased interest in efficient hydrogen storage materials. Hydrogen, as a high-density fuel, offers a green alternative with zero CO2 emissions. Solid-state hydrogen storage technologies, particularly metal hydrides like Magnesium, show potential for safe and high-capacity storage. However, challenges like slow hydrogen kinetics still exist. To address this, researchers are focusing on understanding and optimizing hydrogen diffusion pathways through theoretical modeling and experimental studies. The emergence of Machine Learning accelerated Molecular Dynamics (MLMD) has revolutionized simulations for studying hydrogen-defective systems, enabling accurate predictions over long time scales. By enhancing ML-potentials applications, researchers aim to improve predictions of dynamical properties in metal hydrides without compromising accuracy. The study utilizes fine-tuned potentials and compares them with traditional methods, showing superior performance in predicting the temperature dependence of the diffusion coefficient. This research showcases the potential of machine learning in advancing hydrogen storage technologies and highlights the importance of accurate simulations in developing efficient and economical storage solutions for a sustainable future.