This paper studies responsibility attribution for outcomes of collusive bribery. In an experiment, participants labeled as either citizens or public officials can propose a bribery transaction to another participant (labeled as either public official or citizen, respectively), who decides whether to accept the proposal. We then let either the victims of the corrupt transaction or the bystanders of it judge the individual decisions of proposing and accepting. We interpret these judgments as a measure of responsibility attribution. We find that labels (citizen or public official) have a stronger effect than roles (proposer or responder): public officials are consistently regarded as more responsible for corruption than citizens, while those accepting a bribe are regarded as only somewhat more responsible than those proposing it. Further, we find that victims judge corruption decisions more severely than bystanders, although bystanders' judgments are also consistently negative. In treatments with a neutral context, we find that judgments are less harsh than in the corruption context, bystanders' judgments are much less harsh than those of victims, and responders are judged more severely than proposers. Our results suggest that people judge corrupt actors in context, more harshly when they are labeled as law enforcers (i.e., public officials), and that unaffected parties (i.e., bystanders) react nearly as negatively to corruption as those directly affected by it (i.e., victims).