We present a simple model of the evolution of contracts/rules/organizations via “loopholes.” We study a setting in which the contracting environment is not common knowledge among principals and agents. The economy has a continuum of locations; at each location there is a sequence of principals, each of whom offers a contract to a single agent. Principals offer contracts that deter (at a cost) the behavior they deem to be most likely to harm them. An agent may discover actions that are privately beneficial but are likely to harm her principal, and will undertake them if they are not deterred by the contract (loopholes). Future principals get limited observations of the agents’ behavior, and optimally adjust their own contracts that deter the behaviors they know about, possibly closing the loopholes. The problem is that a loophole-free contract deters all undesired behavior and therefore conveys little information about what actions are feasible. Future principals may then choose loophole-laden contracts. The result is cycling of both undesired behaviors and contracts types. Contracts grow in complexity until they become so costly and the undesired actions perceived as so unlikely that they are then replaced by a simple (loophole-laden) contract.