Achieving a proper representation of storms with large societal impacts requires high spatial resolutions (a few kilometers or finer) to realistically simulate the storm processes (e.g. convection), which is computationally demanding for global modeling. Consequently, downscaling approaches have been used to statistically or dynamically relate storms with their large-scale environments simulated by low-resolution Global Climate Models (GCMs) to project their future changes. The GCM simulated large-scale environmental conditions, however, often display systematic biases, challenging the credible use of downscaling to project future changes of high-impact weather events. In this study, a machine learning (ML) approach was developed and employed to bias correct the large-scale wind, temperature, and humidity simulated by the atmospheric component of the Energy Exascale Earth System Model (E3SM) at ~1^◦ resolution. We demonstrated the usefulness of the ML approach for improving the quality of the low-resolution E3SM simulation of the large-scale environmental conditions associated with three types of high-impact weather systems or storms, including tropical cyclones (TCs), extratropical cyclones (ETCs), and atmospheric rivers (ARs). More specifically, the bias correction directly improves the water vapor transport associated with ARs and the thermodynamical flows associated with ETCs. When the bias-corrected large-scale winds are used to drive a synthetic TC track forecast model over the Atlantic basin, the resulting TC track density agrees better with that of the TC track model driven by observed winds. Applying the ML model to the future climate projections shows that the ML bias correction can preserve the response of the large-scale environments to climate change while effectively reducing the biases in the simulated large-scale states for the historical climate. This study suggests that the ML approach can be used to improve the modeling of extreme weather events by providing more realistic large-scale storm environments simulated by low-resolution climate models for downscaling.