Many RBAC models have augmented the fundamental requirement of a role abstraction with features such as parameterised roles and environment-aware policy. This paper examines the potential for unintentional leakage of information during RBAC policy enforcement, either through the exchange of parameters with external services when checking environmental conditions, or through a policy design which does not appropriately separate policy subsections with different basic purposes. We propose a simple, robust mechanism for handling these problems, and illustrate our approach with a current application of our OASIS RBAC system.