Many experimental studies use random assignment to identify factors influencing dishonesty. However, in real-life, people deliberately choose dishonesty-enabling environments. In two laboratory experiments, we let participants self-select in two tasks, one of which enabled them to cheat. We found that participants low in the honesty-humility were more likely to choose the cheating-enabling task. Furthermore, after choosing it, they cheated even more than when they were randomly assigned to it for the first time. When choosing the cheating-enabling task was costly, the interest in it decreased, but those who chose the task anyway cheated even more. An intervention based on social proof aimed to discourage self-selection into the cheating-enabling environment had the opposite effect. The results suggest that immoral individuals are likely to dominate cheating-enabling environments, where they cheat extensively. Interventions trying to limit the choice of these environments may backfire and lead to the selection of the worst fraudsters.