The aim of this talk is the rigorous derivation of cross-diffusion systems from stochastic, moderately interacting many-particle sys-tems for multiple species. Applications include animal populations and neuronal ensembles. The mean-field limit leads to nonlocal cross-diffusion systems, while the limit of vanishing interaction radius gives local cross-diffusionequations. This allows for the derivation of fluid-type models that can befound in neuronal networks and of Shigesada-Kawasaki-Teramoto populationmodels. The derivation uses the techniques of Oehlschl ̈ager. The entropystructure of the limiting models is discussed and some numerical experimentsare presented.