Wrong-doers often try to collaborate to achieve greater gains than would be possible alone. Yet would-be collaborators face two issues: they need to accurately identify other cheaters and they need to trust that their collaborators do not betray them when the opportunity inevitably presents itself. Unfortunately for would-be dishonest collaborators, these concerns are in tension as the people who are genuine cheaters are likeliest to be untrustworthy. We formalise this interaction in the “villain’s dilemma” and use it in a laboratory experiment to study three questions: what kind of information helps people to overcome the villain’s dilemma? Does the villain’s dilemma promote or hamper cheating relative to individual settings? Who participates in the villain’s dilemma and who is a trustworthy collaborative cheater? We find that public reputation is important for supporting collaborative dishonesty, while more limited sources of information lead to back-stabbing and poor collaboration. We also find that dishonesty is higher in the villain’s dilemma than in our individual honesty settings. Finally, individual factors are generally unrelated to either choosing to collaborate while individual dishonesty predicts whether someone is a trustworthy collaborator or not.