Brain-computer interfaces (BCIs) have been largely developed to allow communication, control, and neurofeed-back in human beings. Despite their great potential, BCIs perform inconsistently across individuals and the neuralprocesses that enable humans to achieve good control remain poorly understood. To address this question, weperformed simultaneous high-density electroencephalographic (EEG) and magnetoencephalographic (MEG) re-cordings in a motor imagery-based BCI training involving a group of healthy subjects. After reconstructing thesignals at the cortical level, we showed that the reinforcement of motor-related activity during the BCI skillacquisition is paralleled by a progressive disconnection of associative areas which were not directly targetedduring the experiments. Notably, these network connectivity changes reflected growing automaticity associatedwith BCI performance and predicted future learning rate. Altogether, ourfindings provide new insights into thelarge-scale cortical organizational mechanisms underlying BCI learning, which have implications for theimprovement of this technology in a broad range of real-life applications.