In this paper, we argue that, while it is necessary to modify existing policy using the lessons learned from disaster events (i.e., reactive learning), this approach is insufficient on its own for dealing with ongoing and emerging climate‐induced disaster risks. Rather, we assert that policymakers must also adopt a proactive and anticipatory learning approach that would enable policy learning and policy evolution in the absence of a major disaster event. We examine drivers, actors, and processes of change in disaster‐management policy paradigms in Bangladesh. A longitudinal learning perspective is applied. We categorize disaster management (DM) policy regimes into three learning episodes: (i) reactive, (ii) transitional, and (iii) proactive. The roles of reactive and proactive learning in shifting DM policy paradigms within these learning episodes are particularly determined. Finally, five interrelated factors that triggered proactive policymaking are identified, which are: risk‐oriented policymaking; cross‐scale (i.e., lesson drawing and policy transfer) and cross‐level (i.e., from local, regional, and national experience) learning; participation of multiple stakeholders; research‐informed and knowledge‐based policymaking; and the presence of a strong advocacy group and a participatory policy process.