When should we expect an ethnic group or more generally opposition group to select an extremist as leader or agent? How does this choice depend on the stakes or value of power sharing and on the presence of corruption? In what way the presence of corruption and extremism and their endogenous interaction determine the likelihood of conflict? We develop a simple bargaining and agency model that yields answers to these questions. We show that when an agent is more likely to be subject to corruption the selection of group leader is more likely to be extreme, hence determining higher likelihood of conflict, which happens precisely when the selected extremist happens to resist the corruption attempts. We further distinguish between political corruption and administrative corruption, providing some nuanced predictions.