Policy decisions often involve a repeated proofreading process before implementation. We present a dynamic model of proofreading decisions by a heterogeneous committee before implementing a risky policy. The proofreading process is necessary because the risky policy contains an unknown number of errors. Proofreading continues as long as a qualified majority votes for continuation. Once the proofreading process ends, and the policy is implemented, members receive heterogeneous penalties based on the remaining errors. We characterize the optimal voting rule given the costs and penalties for the committee. We find that any qualified voting rule, including majority rule, for proofreading is inefficient.