The performance of popular molecular mechanics (MM) force fields in treating problems that involve ion-channel interactions is explored. We have used quantum mechanical/molecular mechanical (QM/MM) calculations to compute the electrostatic potential inside the selectivity filter of the KcsA potassium channel. A comparison is made with the result of classical electrostatic calculations with nonpolarizable MM force fields (AMBER, CHARMM, and GROMOS). An effective procedure is proposed to improve force field charges by performing a fit on the electrostatic potential computed along QM/MM simulations, using a dynamical electrostatic potential derived charge set. The optimized charge set is able to reproduce the QM/MM electrostatic potentials along the channel axis within 1-2 kcal/mol, which represents an improvement relative to the corresponding electrostatic potentials obtained with popular MM force fields. By providing quantum mechanical benchmark charges and energies for the KcsA selectivity filter, we hope to facilitate developments toward the modeling of ion channels by providing an objective test as to whether a given implementation of a new, polarizable, model represents a real improvement over existing fixed point charge models.