Magnetoelectric (ME) materials are of fundamental interest and have been investigated for their broad potential for technological applications. The search for, and eventually the theoretical design of, materials with large ME couplings present challenging issues. First-principles methods have only recently been developed to calculate the full ME response tensor $α$ including both electronic and ionic (i.e., lattice-mediated) contributions. The latter is proportional to both the Born dynamical electric charge $Z^{\rm e}$ and its analogue, the dynamical magnetic charge $Z^{\rm m}$. Here we present a theoretical study of the magnetic charge $Z^{\rm m}$ and the mechanisms that could enhance it. Using first-principles density-functional methods, we calculate the atomic $Z^{\rm m}$ tensors in $\rm{Cr_2O_3}$, a prototypical magnetoelectric, and in KITPite, a fictitious material that has previously been reported to show a strong ME response arising from exchange striction effects. Our results confirm that in $\rm{Cr_2O_3}$, the $Z^{\rm m}$ values and resulting ME responses arise only from spin-orbit coupling (SOC) and are therefore rather weak. In KITPite, by contrast, the exchange striction acting on the non-collinear spin structure induces much $Z^{\rm m}$ values that persist even when SOC is completely absent. ; Comment: 7 pages, 4 figures