The international climate change regime started out very constructively, but although there has been progress over the last 25 years, this progress falls short of what is needed to address the climate change problem. This paper presents the regime evolution in terms of the concept of single-, double-, and triple-loop learning, and its relationship with participatory processes and trust, emphasizing the more recent developments and prospects for the future. It argues that in the first stage of the regime, the problem was seen as structured (high scientific and normative consensus), defined as a technological problem and the focus was on improving the routines within the climate change negotiations. In the second stage, the problem was seen as moderately structured (with creeping doubts about the science and norms needed to deal with the problem) and defined as a political problem requiring double-loop learning that questioned the underlying assumptions. In the third stage, the problem is seen as unstructured (where climate skeptics help shape political discourses on climate science and there is breakdown in normative consensus), as the problem is seen more as an ideological, systemic one requiring complex triple-loop learning (unlearning, transformative learning) in the context of mutual distrust. The issue is whether in this stage we will be able to combine bottom-up initiatives with top-down legally binding goals, Northern and Southern perspectives, to create a global consensus on the need to restructure society such that the problem can once more be seen as structured and manageable.