Active safety systems that start to act moments before the crash might be capable of anticipating the occupant’s position, either by correcting it, or by taking the out‐of‐position into account. To develop such active safety systems, computer simulations of the occupant’s pre‐crash behaviour are very valuable in determining system performance. The objective of this study was to develop a run‐time efficient computer human model that can simulate active as well as passive human behaviour such that it can be used to simulate the pre‐crash and in‐crash phase in one simulation run. The so‐called active human model is a multi‐body model, and is based on earlier developed human models in MADYMO and techniques for controlled active behaviour. The model’s responses in a 1 g car braking, 15 g frontal, 7 g lateral and 3.6 g rear impact were compared to that of the volunteers. It was concluded that the active human model with controlled active behaviour with co‐contraction of the neck muscles better predicts the behaviour of the volunteers than without co‐contraction or completely passive behaviour. With the best fitting co‐contraction levels and reaction times the maximum deviation from the average peak responses of the volunteers was at most 20%.