The assessment of landslide hazard in cold regions (i.e. influenced by snow, glaciers, permafrost) is faced with a twofold challenge: The magnitude of climate change is expected to be greater here that in many other regions, enabling strong shifts in e.g., the probabilities of triggering events such as intense precipitation or snow melt. Additionally, rapid changes such as permafrost degradation, warming firn areas or vanishing glaciers can fundamentally alter system behaviour and thus strongly change its response to a given forcing. As a consequence, the known difficulties of understanding low-frequency high-magnitude events such as landslides are intensified by continued and nonlinear change. Using concrete example from permafrost research, I will discuss general strategies to nevertheless arrive at conclusions that may be of practical relevance. This presentation, of course, heavily draws from the work of collaborators.