Detecting the unexpected threat-relevant stimuli plays a vital role in preschoolers’ daily life safety, but a few studies have investigated how preschoolers process this kind of stimuli. We applied a classical inattentional blindness task (designed by Mack and Rock Inattentional blindness. MIT Press, 1998) to explore whether threat-relevant stimuli could be better detected in an inattentional condition and whether the age and the fluid intelligence could predict the incidence of the detection. With the involvement of two hundred and thirty-nine preschoolers (aged from 4 to 6 years), we found that it was not more likely for preschoolers to detect the threat-relevant stimuli (Knife and Snake) compared with the non-threat-relevant stimuli (Spoon and Snail). The age difference of detection only occurred in the divided attentional condition, but not in the inattentional condition. Moreover, the group of 5-year-old preschoolers with higher fluid intelligence scores was more likely to detect the unexpected stimuli, but the prediction was not powerful. These findings demonstrate that the threat-superiority effect on IB does not occur on preschoolers and the individual difference of preschoolers’ IB is unstable. This study enriches the cognition of young children’s attentional bias to threat-relevant stimuli, and has certain significance to understand the essence of children’s attentional process.