A review of safety targets for en route ATC radar separation suggests that the existing target level of safety (TLS) is over-cautious. If risk budgeting principles are followed consistently, a ‘radar TLS’ of 1·0×10−9 fatal aircraft accidents per flying hour is appropriate. This rate is consistent with Joint Aviation Authorities (JAA) guidance on system failure conditions leading to catastrophic accidents. Dynamic and static calculations using published data are compared. The new methodology shows where there are problems with the traditional static calculations, and how to improve the estimation. A further improvement introduces a simple robust model of the controller's decision processes. The focus is not on describing what controllers would generally do, but on setting criteria based on what they could not reasonably be expected to do. This additional ingredient into the calculation adds realism and ensures that attention is focused on hazardous correlated errors. Focused data collection would be an essential component of new risk estimates. The key information required would be on radar performance and the nature and frequency of use of radar separation, including the relative velocities for proximate events at closest point of approach and the frequency of correlated gross errors (through a conditional probability factor). If this factor is not properly taken into account, then the data collection and analysis could be inefficient.