The ‘human shield hypothesis’ describes the situation where prey species use humans as shield from natural predation. We tested the human shield hypothesis in a population of mountain nyala (Tragelaphus buxtoni) subjected to predation from the nocturnal spotted hyena (Crocuta Crocuta) in the Bale Mountains National Park, Ethiopia by radio-marking 15 mountain nyala (seven females and eight males) and tracking them for up to 2 yr. Occurrence of hyena estimated by faecal transects decreased close to human settlements substantiating the occurrence of a zone with lower risk of hyena predation. The diurnal pattern in the average distance between mountain nyala relocations and human settlements was consistent with the human shield hypothesis with significantly shorter distances during night (when exposed to predation) than during day. However, mountain nyala showed large individual heterogeneity in use of human shields. While nearly all individuals occasionally moved out of the park to human settlements during night, the frequency of such excursions varied from 0% to 71%. The excursions occurred year-round and were not driven by seasonal access to crops. We have previously demonstrated a strong negative effect of humans on the large-scale distribution pattern of mountain nyala. The use of human shield documented here is indicative of a positive small-scale effect of humans. Our study thus supports the view that the effect of human–wildlife interactions can be scale-dependent.