Corruption requires a coalition to form and reach an agreement. Is there a cheap way to stop any agreement from being reached? We find an optimal mechanism that resembles Poker. The players' hands are synthetic asymmetric information, and they create a lemons problem in the market for bribes. Our Poker mechanism is robust: it thwarts bribes regardless of the negotiation procedure, including alternating offers bargaining, Dutch auctions, arbitration, and so on. Our mechanism's cost is inversely proportional to the number of players. So when we embed our mechanism in regulatory approval and regulatory compliance settings, we find that it is optimal to hire competing auditors to each case. In compliance cases, there is a trade-off between rewarding the agent for honesty and punishing the agent for non-compliance. This trade-off is resolved by rigging the Poker hand distribution against the agent. Our results are also pertinent to robust mechanism design and worst-case information structures.