Within a standard three-tier regulatory model, a benevolent principal delegates to a regulatory agency two tasks: the supervision of the firm's (two-type) costs and the arrangement of a pricing mechanism. The agency may have an incentive to manipulate information to the principal to share the gains of collusion with the firm. The novelty of this paper is that both the regulatory mechanism and the side contracting between the agency and the firm are modelled as a bargaining process. While as usual the inefficient firm does not have any interest in cost manipulation, we find that the efficient firm has an incentive to collude only if the agency's bargaining power is high enough, and the total gains of collusion are now lower than those the two partners would appropriate if the agency could make a take-it-or-leave-it offer. Then, we focus on the optimal institutional responses to the possibility of collusion. In our setting, where the incompleteness of contracts prevents the principal from designing of a screening mechanism and thus Tirole's equivalence principle does not apply, we show how the players' bargaining powers crucially drive the optimal response to collusion.