A novel quantitative approach to identify the hazards associated with a process is presented. It is based on the analysis of a "region-transition model" of the process which, through the concepts of hybrid state transition systems and interval arithmetic, accounts for both nonlinearity and uncertainty. A safety verification algorithm which evolves the model over a. given time horizon is able to determine which hazards can occur. As required in safety analysis, the methodology always provides a conservative appraisal of the dangers, based on the best model available. An exothermic reaction taking place in a batch reactor serves as an illustrative example.