This study presents a driving simulator experiment conducted on 47 drivers to investigate how different populations of users respond to automated system failure. This study is mainly focused on examining the behavioral trends of conservative users versus aggressive users. On this account, a major takeover scenario of a level 3 automated vehicle malfunctioning at three high-speed critical curves along a freeway was designed. The drivers are notified with an auditory warning that is triggered instantaneously with the malfunctions, thus indicating a demand to takeover. While riding in autopilot mode, the participants were instructed to use their phones to browse and watch YouTube, to allow them to become comfortable with automation. The reaction time, time to regain control, frequency of time to regain control, frequency of unsafe curves, and type of control were used as measures of users’ behavior. The results show that while some measures are not different between the conservative and aggressive users, other measures indicate that conservative users tend to learn as they experience more malfunctions and adjust their response to improve safety. However, there is enough evidence that such group of users are more likely to drop their level of trust in automation if they experience unsafe maneuvers or lose control. These findings are promising as they can help auto-makers better design autonomous vehicles and officials better establish educational programs, which can accommodate different groups of users.