Invoking memories and imagery from the Holocaust and other German atrocities during World War II, many contemporary commentators and politicians believe that the international community has an affirmative obligation to deter and incapacitate perpetrators of humanitarian atrocities. Today, the received wisdom is that a legalistic approach, which combines humanitarian interventions with international criminal prosecutions targeting perpetrators, will help realize the post-World War II vision of making atrocities a crime of the past. This Article argues, in contrast, that humanitarian interventions are often likely to create unintended, and sometimes perverse, incentives among both the victims and perpetrators of atrocities. The problem is that when the international community intervenes in the civil wars or insurrections where most humanitarian atrocities take place, its decision is partially endogenous or interdependent with that of the combatants; humanitarian interventions both influence and are influenced by the decisions of the victims and perpetrators of atrocities. Herein lies the paradox: because humanitarian interventions tend to increase the chance that rebel or victim group leaders are going to achieve their preferred political objectives, such leaders might have an incentive to engage in the kinds of provocative actions that make atrocities against their followers more likely in the first place. More specifically, the prospect of humanitarian intervention often increases the level of uncertainty about the distribution of costs and resolve between the combatants. In turn, such uncertainty amplifies the possibility of divergent expectations between the dominant and rebel group regarding the outcome of a civil war. At bottom, the prospect of humanitarian intervention might sometimes increase the risks of genocidal violence. This Article turns to insights from the domestic framework of torts and criminal law to elaborate upon the theoretical framework that motivates this perverse dynamic, provides some contemporary illustrations from civil wars in Africa and the Balkans, and recommends improvements to the current regime to mitigate some of its unintended effects. This Article concludes that the optimal regime of humanitarian intervention would incorporate comparative fault principles that take into account the failure of victim (or rebel) leaders to take adequate precautions against the risks of humanitarian atrocities.