Agile Safety Case Assessments for Autonomous Vehicle Fleets

The large-scale deployment of autonomous vehicles (AVs) demands not only robust safety cases but also systematic approaches for their assessment. A safety case is a structured argument, supported by evidence, that a system is acceptably safe within its intended domain. While many AV developers are now adopting safety cases as a prerequisite for launch readiness, effective assessment methods have not kept pace with the field’s rapid progress. This paper presents a three-stage methodology for safety case assessment—framework, solution, and implementation assessment—developed through nearly fifty evaluations across commercial and defense AV programs. Framework assessment benefits from leveraging safety case frameworks, which accelerate development and prevent argument defects. Solution assessment focuses on defining clear evidence criteria, while implementation assessment evaluates the sufficiency of complete evidence packages. Together, these stages support iterative improvement and establish traceability to emerging Automotive Vehicle Safety Consortium (AVSC) best practices for safety case evaluation. We further highlight the growing role of Safety of the Intended Functionality (SOTIF) in assessment. Building on recent research with the SOTIF Meta-Algorithm, we show how structured hazard analysis, quantitative fault trees, and Safety Performance Indicators (SPIs) can provide empirical grounding for claims. Data-driven techniques, such as distributional shift detection between simulation and real-world metrics, extend the assurance process to capture unknown unsafe scenarios and edge cases. By integrating safety case frameworks with AVSC best practices and SOTIF-informed analytics, we outline a path toward continuous and data-driven safety case assessment. This combined approach enables earlier detection of gaps, reduces costly rework, and increases confidence in safety readiness—ultimately helping AV developers and assessors establish credible, resilient assurance processes aligned with evolving regulatory and societal expectations.