This paper examines the effect of political power on the risk of legal prosecution. We focus on the case of India for which the most comprehensive data are available. Over 40% of India’s legislators face criminal charges. However, ethnographic evidence suggests that those who get the charges, are not necessarily the actual criminals. Rather, criminal charges seem to be endogenous to political power. We test this hypothesis on the basis of a dataset of winning candidates and their closest competitors in the 2009 national election and follow the development of their charges over five years until the next election. Our results reveal that political power indeed matters, and suggest a variety of channels for this effect. Incumbency power of MPs is particularly relevant in southern states. Using a regression discontinuity design, we show that until the 2014 election, incumbents cumulate on average two charges less than those politicians that narrowly lost the election. In other states, incumbency advantages appear to be outweighed by other political factors such as alignment with the ruling party. Across the whole country, affiliation with a strong national or state party (as opposed to independent or unrecognized local parties) reduces criminal charges. These results reveal a serious and systematic governance problem. They also reveal that caution is required when interpreting reported criminal charges at the level of individual politicians. A better understanding of who are the actual criminals is required to solve the problems induced by criminality in office.