Modern society heavily relies on strongly connected socio-technical systems. As a result, distinct risks threatening the operation of individual systems can no longer be treated in isolation. Risk experts are actively seeking ways to relax the risk independence assumption that undermines typical risk management models. Prominent work has advocated the use of risk networks as a way forward. However, the inevitable biases introduced during the generation of these survey-based risk networks limit our ability to examine their topology and in turn challenge the utility of the very notion of a risk network. To alleviate these concerns, we propose an alternative methodology for generating weighted risk networks. We subsequently apply this methodology to an empirical data set of financial data. This paper reports our findings on the study of the topology of the resulting risk network. We observe a modular topology and reason on its use as a robust risk classification framework. Using these modules, we highlight a tendency of specialization during the risk identification process, with some firms being solely focused on a subset of the available risk classes. Finally, we consider the independent and systemic impact of some risks and attribute possible mismatches to their emerging nature.