We study the shareholder-manager relationship when a fraudulent strategy is available. In a canonical agency setting, we introduce privately known CEO’s morality, modeled as a cost of breaking the law. We derive the optimal compensation offered by the firm and examine how it affects the CEO’s action. In the optimal contract, there are two regimes, depending on the quality of law enforcement : providing incentives and preventing fraud can be either complements or substitutes. As a consequence, either the variable or the fixed part of remuneration helps preventing fraud. We also point out that, given the agency problem, the level of corporate fines cannot be a substitute for low levels of detection. Finally, the comparative statics of our model shed light on contradictory empirical evidence in the literature.