In this talk, I will review the observational evidence for an intimate connection between the baryonic surface density and the total gravitational field in disk galaxies. This observational fact presents a fine-tuning problem for the particle dark matter interpretation of mass discrepancies in galaxies. I particular, it leads to a too high "diversity" of rotation curve shapes for a given mass. On the other hand, this phenomenology is naturally explained in a paradigm hypothesizing an effective breakdown of Newtonian dynamics in the extremely low acceleration regime (MOND). However, MOND predictions break down on scales larger than galaxies. Theories modifying the lagrangian of the dark matter sector to account for the observed phenomenology in galaxies while preserving the predictions of the standard cosmological model on the largest scales are perhaps a promising way to reconcile these conflicting observational facts. I will present some tentative dark matter models which seem to fit the bill, as well as their current shortcomings.