The rational design of ionic liquid electrolytes (ILs) is hindered by a persistent trade-off in molecular simulation: classical force fields offer computational efficiency but often require laborious, system-specific parameterization and may exhibit limited transferability across chemical environments, whereas first-principles methods provide high accuracy but cannot readily access the length and time scales required to capture the slow, collective dynamics governing ion transport. Machine-learning interatomic potentials offer a promising route to bridge this gap by approaching density functional theory (DFT) accuracy at substantially reduced computational cost. Here, we demonstrate that a moment tensor potential (MTP), trained on DFT reference data through on-the-fly active learning, enables accurate and computationally efficient simulations of diluent-modulated ILs over extended time scales. By selectively incorporating informative configurations into the training set, active learning substantially reduces the amount of reference data required while maintaining accuracy across a broad range of diluent concentrations. The resulting MTP reproduces key structural and thermodynamic properties, including lattice parameters and densities, in close agreement with DFT calculations. More importantly, long-timescale simulations capture the concentration-dependent evolution of solvation structures and ionic transport upon dilution, consistent with previous computational studies and experimental measurements. The MTP thus reproduces not only equilibrium properties but also the concentration-dependent trends characteristic of locally concentrated ionic liquids, where changes in local solvation environments strongly influence ion-transport pathways. These results demonstrate that actively learned MTPs can combine first-principles-level accuracy with transferability across composition and access to the time and length scales required to describe ion transport. This framework provides a practical route toward predictive, first-principles-quality simulations for the rational design of ionic liquid electrolytes.