Throughout the world, states delegate various and sundry security tasks to armed groups outside their regular forces. The main arguments in the literature are that governments use militias because of their logistic advantages (autonomy, informational advantages, and cheap deployment) and to provide the government with plausible deniability for human rights violations. However, not all militias victimize civilians and when they do the government usually does as well. Why, then, would states use militias at all? Moreover, why, despite the risks of side-switching, agency loss, and excessive use of force inherent to them, do states use armed groups outside their regular forces rather than spending more on the latter? This study develops a general framework to analyze the use of militias. It presents a formal model of the state's decision to spend on regular and irregular forces as if it was a firm producing a good: security. However, these forces also produce liabilities to the state, and so it is in balancing security and liability that the state optimizes its expenditures. The theory leads us to expect that militias are more likely to be used, all else equal, the greater the budget available is, the more cost-effective they are relative to the regular forces, the lower the risk of using them, and the more a state values security over liability. Likewise, there is substitutability between armed forces: that to spend on one force the state must necessarily spend less on the other one. Thus, all else equal, the more a state uses militias, the less it will use its regular forces, and vice-versa. I test some of these expectations with two studies. The first uses cross-national data on militias, judicial independence and civil wars between 1981 and 2007. I find that militias are more likely to be used in states that have a weak judiciary. The second is a case-study of Colombia's most recent civil war. Using municipality-level data between 2000 and 2006, I find that the Colombian armed forces decreased their activity against rebels in areas where the United Self-Defense Forces of Colombia were active.