DescriptionIn this dissertation, we provide a theory of time-consistent dynamic risk measures for partially observable Markov jump processes in continuous time. By introducing risk filters, which are new two-stage risk measures, we show that the risk measure of a partially observable system can be represented as a risk measure of a fully observable system that is defined by a g-evaluation. The innovation process of the original system is the Brownian Motion driving the new fully observable system. Furthermore, we introduce a risk-averse control problem for the partially observable system and we derive a risk-averse dynamic programming equation.