Previous work has shown that drivers engaged in concurrent cognitive tasks exhibit some adaptive behaviors to enhance safety, such as increasing their headway distance, despite the fact that other aspects of safety might be compromised. However, these studies often test steady-state behaviors (e.g., car-following) that might not be representative of traffic situations in which drivers' goals and intentions are constantly changing (i.e., involving tactical vehicle control). In two simulator experiments, we examined the impact of mental workload on drivers engaged in a “naturalistic” tactical driving task. In Experiment 1, we explored the safety margins (distances) that drivers maintain between themselves and vehicles around them when completing a passing maneuver. In Experiment 2, we examined safety margins and performance under less constrained, yet more realistic and dynamic conditions. In both experiments, we found no evidence that drivers adjust their safety margins to account for the additional demands of performing a cognitive task. The implications for steady-state experimental scenarios versus more dynamic ones are discussed.