Instead of seeing accidents traditionally as the result of event-chains, STAMP defines an accident as an inadequate implementation of safety constraints within the system's structure. The causal factors of accidents lie within the differing mental models about the system's structure and behaviour of the system's controllers. Policies and system's safety are interwoven. Despite unquestionable merits, traditional models in engineering and political science are note capable of analyzing the challenges deriving from the complexity of systems as well as policy processes. Nevertheless, the adaption of cybernetics to political science was fruitful at due time. Therefore, we combined a modern hazard analyzing methodology (STAMP) with a recent perspective in policy analysis. Our results show that MSP is capable of analyzing policy processes under conditions of ambiguity. Actors involved in ambiguous processes have problematic preferences; their mental models might be hampered by a broad variety of ideas and other cognitions referring to the same phenomenon in a system. Moreover, policy entrepreneurs might push a pet policy that impairs system's safety. The Multiple Streams Perspective helps to understand those ambiguous policy processes. Nevertheless, the MSP is not able to quantify policies in order to get a holistic view on system's safety. The newly developed Cybernetic-oriented Modeling of Policy Processes (COMPP) exemplifies that ambiguous policy processes can be integrated in hierarchical safety control structures. Although human behavior generally lacks of a predictable 'if-then' function, policy makers and policy entrepreneurs can be conceptualized as automated respectively human controllers within a generic control loop. Both of them have sensors to measure specific parts of the operative processes. However, neither of them has direct actuators to influence the policy process. Within a hierarchical Safety Control Structure the problem-, policy- and politics stream can be understood as independent control components of the system. Each stream is connected to different other control components. Thus, our results support our notion that formerly qualitative models of policy analysis can be quantified by combining them with hazard analyzing methodologies. Further research should apply our model to different case studies and a system dynamics modeling in order to identify strengths and weaknesses of our hybrid methodology.