Global challenges like the climate crisis and pandemic outbreaks require collective responses that quickly adapt to changing circumstances. Social norms are potential solutions, but only if they are capable of adapting themselves. Despite a large literature showing the potential of social norms to promote cooperation in collective action problems, such as energy conservation, vaccine uptake or tax compliance, less is known about how social norms themselves are affected by the changing context, potentially compromising their effectiveness in solving dilemmas. In this talk, I will discuss the results of four recent studies, a long-term experiment, a replication of this long-term experiment conducted during the first wave of the Covid-19 pandemic, an agent-based simulation to study the dynamics of social norms under collective risk and a long-term survey measuring social norms of distancing, how they change over time and their effect in promoting cooperation under risk. Taken together results from these studies show that it is not obvious that norms are effective solutions to deal with (changing) collective risk, since the norms themselves are affected by risk as well. If we want to leverage social norms to promote long-lasting cooperation, we need a better understanding of the feedback cycle between norms, risk, and behavior.