Whistle-blowing is seen as a powerful tool in containing corruption, although theoretical findings and experimental evidence cast doubt on its effectiveness. We expand a standard corruption model by allowing both, briber and official to initiate corruption actively, in order to assess the full effect of whistle-blowing. In our laboratory experiment we find that the effect of symmetrically punished whistle-blowing is ambiguous since it reduces the impact of corruption on productive activity, but also increases its stability. We show that asymmetric leniency for the official offsets the negative effect. The results can be explained by simple arguments about belief structures within the self-interested model of payoff maximizing.