Administrative rules touch on almost every aspect of Americans’ lives, from the fuel standards in our cars to whether the “Plan B” morning-after pill is sold at the local pharmacy. Yet in spite of this importance, scholars have very little understanding of how politics affects whether agencies set policy moderately or extremely through rulemaking. In this paper, I propose a model where an agency sets a proposed rule based on the preferences of Congress and the president and subsequently adjusts that policy based on feedback from interest groups. The core insight of the model is that agencies leverage interest group feedback on rulemaking proposals to wrest policy concessions from their political overseers. I test the empirical implications of the theory using innovative text analysis techniques on a new dataset more than 1,000 agency rules. The project speaks to the democratic accountability of the American bureaucracy and whether additional oversight is warranted.