A series of Taylor rod-on-anvil experiments have been performed to validate the predicted impact velocity fracture threshold and fracture mode of a high hardness armour steel (HHA) obtained through explicit finite element simulations. Experimentally, the rods exhibited principal shear failure, a condition that can be closely linked to adiabatic shear band (ASB) formation in high strength steel. Using a stress triaxiality and Lode angle dependent failure strain criterion (Basaran 3D fracture locus), calibrated from quasi-static mechanical characterisation tests, the simulations were unable to predict the onset of fracture observed in experiments. As such, a strength-fading criterion is proposed using a phenomenological description to capture the loss of load-carrying capacity resulting from ASB formation. The ASB criterion is based on an exponential fit to experimentally-observed instability strains measured at different average stress triaxialities in a series of tests on inclined cylindrical and modified flat-hat specimens. With the prediction of ASB formation the material strength is reduced to model the thermal softening experienced in the shear band, and fracture of the material (in the form of element erosion) remains controlled by the Basaran fracture model. Incorporating the ASB-based criterion, the numerical models were found to accurately predict both the impact velocity fracture threshold, as well as the general appearance of the observed principal shear fracture. The proposed criterion enables the effects of ASB formation to be captured in an impact simulation with little increase in computational cost.