Towards LLM-augmented Situation Space Analysis for the Hazard and Risk Assessment of Automotive Systems

The hazard analysis and risk assessment (HARA) is a fundamental artifact of safety engineering that requires substantial effort from experienced engineers to conform to standards. The cost combined with today’s shortage of trained personnel demands technical solutions to reduce these efforts. Recent advances in generative AI provide new opportunities to support engineers with complex and creative tasks such as the HARA that require knowledge of the system and the environment. However, for now, the results provided by large language models are not reliable enough for safety artifacts so human engineers must review and approve them. In our study, we present a tool to create a HARA based on human-AI cooperation. Our goal is to keep human engineers involved in each decision and guide them by providing arguments, examples, and assessment proposals. In our case study, the engineers benefit from the additional reasonings and proposed assessments of the AI tool.