The Necessity of a Holistic Safety Evaluation Framework for AI-Based Automation Features

The intersection of Safety of Intended Functionality (SOTIF) and Functional Safety (FuSa) analysis for driving automation features has traditionally excluded Quality Management (QM) components from rigorous safety impact evaluations. Historically, QM components have not been classified as safety-relevant, and therefore their potential influence on overall system safety has often been overlooked. However, recent advancements in artificial intelligence (AI) integration within automated driving systems have revealed that these components can, under certain conditions, contribute to SOTIF-related hazardous risks. This emerging reality challenges long-standing assumptions and underscores the need for a more inclusive approach to safety analysis. Furthermore, compliance with evolving AI safety standards, such as ISO/PAS 8800, necessitates a re-evaluation of safety considerations for components previously deemed outside the scope of FuSa and SOTIF frameworks. This paper examines the necessity of conducting holistic safety analysis and comprehensive risk assessment on AI components, emphasizing their potential to introduce hazards capable of violating established risk acceptance criteria when deployed in safety-critical driving systems. Particular attention is given to perception algorithms, where deficiencies can lead to unintended functional behaviors with significant safety implications. Through the use of case studies, we demonstrate how such deficiencies may arise even in QM-classified components, illustrating the practical consequences of insufficient safety oversight. By bridging theoretical analysis with real-world examples, this paper argues for the adoption of integrated methodologies that combine FuSa, SOTIF, and AI standards-driven practices to identify, evaluate, and mitigate risks associated with AI components. The findings highlight the importance of revising existing safety frameworks to address the evolving challenges posed by AI technologies, ensuring comprehensive safety assurance across all component classifications and aligning with multiple safety standards.