Current climate policy formulations largely ignore our limited understanding of Earth system processes and potential feedbacks. It is standard practice to design climate policies under a well-behaved Earth system and then estimate the probability of their climate outcomes (ex-post). Robustness of climate policies, however, requires incorporating physical uncertainty at the design stage (ex-ante). Currently, robust policies are not discussed on the international stage, as they have remained confined to stylised academic studies. Here, we employ a novel integrated assessment framework that embeds state-of-the-art estimates of physical uncertainty, obtained through Bayesian fusion of the latest data from Earth system models and observations, to derive robust global climate mitigation strategies. Compared to their non-robust counterparts, robust strategies call for precautionary measures. Under a variety of cost-benefit and cost-effective experiments, net-zero CO2 emissions must typically be reached a decade earlier, which requires23
paying a near-term precautionary premium that can more than double the carbon price. In the long term, the precautionary measures entail developing and sustaining negative emission technologies for centuries. Beyond demonstrating a radical paradigm shift with the ex-ante approach, our work reassesses upward the challenge faced by humanity to build a robustly safe future within Earth system boundaries.