We investigate the energy loss characteristics of warm dense matter (WDM) and dense plasmas concentrating on the influence of electronic correlations. The basis for our analysis is a recently developed ab initio Quantum Monte-Carlo (QMC) based machine-learning representation of the static local field correction (LFC) [Dornheim et al., J. Chem. Phys. 151, 194104 (2019)], which provides an accurate description of the dynamical density response function of the electron gas at the considered parameters. We focus on the polarization-induced stopping power due to free electrons, the friction function, and the straggling rate. In addition, we compute the friction coefficient which constitutes a key quantity for the adequate Langevin dynamics simulation of ions. Considering typical experimental WDM parameters with partially degenerate electrons, we find that the friction coefficient is of the order of γ/ωpi = 0.01, where ωpi is the ionic plasma frequency. This analysis is performed by comparing QMC based data to results from the random phase approximation (RPA), the Mermin dielectric function, and the Singwi-Tosi-Land-Sjölander (STLS) approximation. It is revealed that the widely used relaxation time approximation (Mermin dielectric function) has severe limitations regarding the description of the energy loss properties of correlated partially degenerate electrons. Moreover, by comparing QMC based data with the results obtained using STLS, we find that energy loss properties are not sensitive to the inaccuracy of the static LFC at large wave numbers k/kF > 2 (with kF being the usual Fermi wave number), but that a correct description of the static LFC at k/kF 1.5 is important.