The growing use of computers to control critical functions of complex systems brings with it concerns over dependability of the software. Engineers involved in developing such systems use a range of techniques for hazard and risk assessment, each offering a different safety perspective. Confidence in the system, and ultimately its certification, is (partially) dependent on confidence in consistency and traceability between results of the various analyses, and between each set of results and the system design. That is, software safety may be viewed from one perspective as an information management problem. Clearly, tool support is essential. However, current tool-sets are incapable of achieving consistency at the fine level of granularity required to justify such confidence. This paper presents one approach towards achieving that objective.