A Framework to Characterise Dynamic Safety Assurance

The concept of dynamic assurance has been discussed for a number of years, often aligned with dynamic safety cases. In broad terms, it refers to ongoing updates to assurance data and the safety case during operation of a safety critical system. In recent years there has been an impetus towards its use with the increase in very complex systems, systems of systems, and AI-enabled safety critical systems. This is due to the inherent difficulty in assur- ing these prior to deployment, leading to a proposed shift to assurance gathering processes during operation. However, there is a large difference between periodic updates of evidence, e.g., from monitoring performance data, and adapting to changes in risk profiles on the fly, e.g., new hazards emerging. This paper aims to clarify different concepts and degrees of dynamic as- surance and dynamic safety cases, including changes to evidence, safety claims, risks, and frequency of change. We use three examples to illus- trate the different degrees and complexity of dynamic assurance and updates the safety case. We argue this type of framework is important as it helps developers of safety critical systems understand the challenges they will face when implementing dynamic assurance approaches, includ- ing the safety case. Additionally it will help reviewers such as regulators, who may need to change practice and guidelines to accommodate more dynamic safety practices. We present a series of challenges dynamic as- surance will bring to safety practitioners and researchers