Utilizing an expanded version of the perceptual load model, we conducted a driving simulator study to test the influence of a Collision Avoidance Warning System on drivers’ performance. Perceptual load on the road (e.g., vehicle congestion) and on the roadside (e.g., number of pedestrians) were manipulated, and critical events were simulated on the road (e.g., a leading car suddenly slowed down) or initiated from the roadside (e.g., a pedestrian crossed the road unexpectedly). Each participant drove in four different scenarios: two using the warning system and two without. We found that in at least one condition (low percptual load levels in both regions) the system served as a two-edged sword. On the one hand, it decreased collisions with vehicles on the road, but on the other hand, increased collisions with elements arriving from the roadsides. These findings demonstrate the importance of systematic manipulation of perceptual load across the visual field and of the location critical events when evaluating drivers’ behavior.