Why does bureaucratic corruption occur in the EU system? Several examples suggest that bureaucratic corruption exists and that the Commission's anti-fraud agency, OLAF, is not a fully independent authority. We thus develop a novel interpretation of the principal-supervisor-agent model to cope with non-independent anti-fraud units. This model shows that corruption is likely to occur when the expected value to the client from bribing the agent is larger than the expected value to the principal of truth-telling by the supervisor. Overall, this analysis points to the risks of flawed incentives and the lack of institutional independence among principal, agent, supervisor and client. Our main policy recommendations as a result of these findings are that OLAF should be placed outside the Commission, and that whistleblowers should receive adequate protection.