This paper looks at the current practices in regional and portfolio seismic risk assessment, discusses some of their shortcomings and presents proposals for improving the state-of-the-practice in the future. Both scenario-based and probabilistic risk assessment are addressed, and modelling practices in the hazard, fragility/ vulnerability and exposure components are presented and critiqued. The subsequent recommendations for improvements to the practice and necessary future research are mainly focused on treatment and propagation of uncertainties.