Looking at functional safety of vehicles, we have seen an evolution from federated to integrated E/E architectures. When extending the way of specifying and analysing functional safety to also address cooperative functionality, it is not possible to keep a static view of the boundaries of the system for which to ensure safety. This is because the set of vehicles realizing a cooperative function may change a lot during the execution of the cooperative function. In this work in progress paper we suggest to move part of the task to show safety, from design time to run time. This implies that it will become necessary to monitor the system at run time, continuously calculate its quality and share that information between the individual vehicles to assert that the system is safe. In order to accomplish this, appropriate metrics are needed, both during design time and run time. Inspired by information theory, this paper sketches some common properties for metrics, and indicates how that can be beneficial.