The density of the Earth’s inner core is lower than that of pure iron, indicating the presence of light elements. Among the candidate elements, carbon, hydrogen, oxygen, sulfur, and silicon are considered the most likely. Viscosity is a key physical property controlling the dynamics and evolutionary history of the inner core, and it has significant implications for the origin of seismic anisotropy. Previous studies have investigated the viscosity of pure iron in its hexagonal close-packed (HCP) and body-centered cubic (BCC) phases under inner-core conditions through computational simulations. However, the influence of light elements on the viscosity of the inner core remains insufficiently constrained. In this study, we constructed a neural network potential (NNP) for Fe-S alloy under inner-core conditions and employed it to perform large-scale molecular dynamics simulations. We systematically examined the impact of vacancy concentrations as low as 0.01% on the ionic transport properties of Fe-S alloy. Based on the self-diffusion coefficients of Fe in the lattice, we further explored the creep mechanisms and viscosity of Fe-S alloy under core conditions. Our results indicate that dislocation creep dominates the rheological behavior, yielding viscosities of 1×10¹⁴–2×10¹⁶ Pa·s, consistent with constraints from free-core nutation and seismic observations.