In this paper we characterize the preferences of a pessimistic social planner concerned with the potential costs of extreme, low-probability climate events. This pessimistic attitude is represented by a recursive optimization criterion à la Hansen and Sargent (1995) that introduces supplementary curvature in the social preferences of standard linear-quadratic optimization analysis and, under certain conditions, it can be shown to correspond to the Epstein-Zin recursive utility. The introduction of extra convexity and the separation between risk-aversion and time-preference implies that, independently of the choice of the discount rate, a sharp, early and steady mitigation effort arises as the optimal climate policy, supporting the main recommendation of the Stern Review (Stern, 2007). Nonetheless, we accommodate for its main criticism of using a too low and questionable discount rate (Nordhaus, 2007), while preserving the assumption of a normal (thin-tailed) probability distribution (Weitzman, 2009). Finally, we argue that our theoretical framework is sufficiently general and robust to possible mis-specifications of the model.