In the past decades, risk assessment frameworks have been developed as a viable alternative to Security-Constrained Dispatch for power grids. Risk-based dispatch can increase power grid profitability and help better understanding of prominent hazards, however, relevant sources of uncertainty are often neglected to meet efficiency requirements. In this work, a research effort towards the development of advanced risk assessment frameworks for power grids is presented. A surrogate model strategy is proposed and speeds up the post-contingency consequence evaluation by analytical sensitivity indices (Line-Outage Distribution Factor). The risk surrogate model is embedded within a generalized uncertainty quantification framework to evaluate: (1) The amount of epistemic uncertainty affecting the risk index and attributable to imprecision in the parameters of the model; (2) The effect aleatory uncertainty in environmental-operational variables. The surrogate model is tested on the 24 nodes IEEE Reliability Test System. Computational time and accuracy of the results are compared to a high-fidelity AC power flow risk model. Strengths and weaknesses of the proposed framework highlighted.