This study investigates the reduction process of oxide-derived copper (OD-Cu) catalysts for electrochemical carbon dioxide (CO₂) reduction. Using large-scale molecular dynamics simulations with a neural network potential trained on first-principles data, the researchers explore the evolution of OD-Cu structures and the distribution of oxygen under different conditions. The results show that the oxygen concentration in OD-Cu increases with pH, potential, or specific surface area. While OD-Cu fully reduces to copper in long electrochemical experiments, removing all trapped oxygen takes a significant amount of time. The highly reconstructed copper surface provides various sites for oxygen adsorption, but surface oxygen atoms are not stable under common experimental conditions. The study provides insights into the dynamics of OD-Cu catalysts and residual oxygen during the reaction, contributing to a deeper understanding of the active sites.This study investigates the reduction process of oxide-derived copper (OD-Cu) catalysts for electrochemical carbon dioxide (CO₂) reduction. Using large-scale molecular dynamics simulations with a neural network potential trained on first-principles data, the researchers explore the evolution of OD-Cu structures and the distribution of oxygen under different conditions. The results show that the oxygen concentration in OD-Cu increases with pH, potential, or specific surface area. While OD-Cu fully reduces to copper in long electrochemical experiments, removing all trapped oxygen takes a significant amount of time. The highly reconstructed copper surface provides various sites for oxygen adsorption, but surface oxygen atoms are not stable under common experimental conditions. The study provides insights into the dynamics of OD-Cu catalysts and residual oxygen during the reaction, contributing to a deeper understanding of the active sites.