Metal closo-hydroborates, particularly Na 2 B 12 H 12 , have emerged as promising solid-state electrolytes for sodium-ion batteries, owing to their high ionic conductivity at elevated temperatures. Despite their potential, the mechanisms underpinning their superionic conduction and phase transitions remain incompletely understood. In this study, we develop a machine learning force field (ML-FF) for Na 2 B 12 H 12 , enabling large-scale molecular dynamics simulations that capture the intricacies of its phase transition, anion reorientation dynamics, and ionic conductivity. Our simulations reveal a martensitic transformation from a monoclinic to a body-centered cubic (bcc) phase at ∼675 K, resulting in a dramatic enhancement in Na + conductivity and a significant decrease in the activation energy for anion reorientation. Additionally, cooling simulations indicate an incomplete reverse phase transition due to rapid cooling, highlighting the complexities of phase stability in Na 2 B 12 H 12 . These findings emphasize the critical role of anion reorientation and the cooperative movement of cations and anions in facilitating superionic conduction. By leveraging the accuracy and efficiency of ML-FFs, this study provides unprecedented atomistic insights into the mechanisms driving high-temperature conductivity in Na 2 B 12 H 12 , offering pathways for the optimization of closo-hydroborates as next-generation solid-state electrolytes.