In May 2000, E. coli originating from nearby agricultural lands contaminated the municipal water supply of Walkerton, Ontario. As a result, over two thousand people became seriously ill and seven people lost their lives. In response to this crisis, source water protection emerged as part of a multi-barrier approach for the provision of safe drinking water. Intervention at the source provides an early opportunity to contain a range of potential risks, many of them tied to land-use. However, source water protection involved a fundamental shift in Ontario’s policy approach to the provision of safe drinking water. In doing so, it mobilized powerful actors to defend their interests against this change. This study traces how the problem was defined in Ontario, and by whom, establishing a continuum of actor–institution interactions that spans the development and implementation stages of the Clean Water Act (2006), and shows how different preferences were carried forward through the devolution of decision making to the watershed level. By disaggregating the policy change into its constituent parts, and accounting for actor effects at the implementation stage, we observe that decentralization in the context of sustained political pressure led to an effective concentration of decision-making power, thereby actually eroding local control. Caution is thus warranted when considering the devolution of decision making to inclusive social processes, as this may link policy subsystems and thereby create the institutional channels through which special interests can dominate decision making.