This paper presents a brief summary of the methodology used for the assessment of risk arising from fuel handling and dismantling operations in advanced gas-cooled reactor power stations. The difficulties with and problems arising from such risk assessments are discussed. In particular, difficulties arise from (i) the onerous risk criteria that nuclear plants are expected to satisfy, (ii) the necessary complexity of the plant, (iii) the conflicting requirements for the fault consequence assessments to be bounding but not grossly pessimistic, and (iv) areas of fault frequency assessment which contain possibly subjective considerations such as software and common mode failure.