We show that standard density functional theory leads to large errors in the electron density distribution compared to reference second order Møller-Plesset perturbation theory (MP2) calculations for the insulin molecule and zwitterionic peptides, while range-separated versions perform much better. The error is quantified in terms of the electrostatic potential (ESP) on a molecular surface, which shows that standard density functional theory incorrectly predicts partial electron transfer from anionic to cationic sites. In addition, we compare the MP2 calculated ESPs to those predicted by commonly used force fields. Several fixed charge force fields display very similar performances with rather large errors, while polarizable force fields significantly reduce the error. Solvation enhances the molecular ESP, which is partly accounted for by fixed charge force fields, but polarizable force fields again perform significantly better.