University of California Institute of Transportation Studies

Vehicle automation represents a new safety frontier that may necessitate a repositioning of our safety oversight systems. This white paper serves as a primer on the technical and legal landscape of automated driving system (ADS) safety. It introduces the latest AI and machine learning techniques that enable ADS functionality. The paper also explores the definitions of safety from the perspectives of standards-setting organizations, federal and state regulations, and legal disciplines. The paper identifies key policy options building on topics raised in the White House’s Blueprint for an AI Bill of Rights, outlining a Blueprint for ADS safety. The analysis concludes that potential ADS safety reforms might include either reform of the Federal Motor Vehicle Safety Standards (FMVSS), or a more holistic risk analysis “safety case” approach. The analysis also looks at caselaw on liability in robotics, as well as judicial activity on consumer and commercial privacy, recognizing that the era of AI will reshape liability frameworks, and data collection must carefully consider how to build in accountability and protect the privacy of consumers and organizations. Lastly, this analysis highlights the need for policies addressing human-machine interaction issues, focusing on guidelines for safety drivers and remote operators. In conclusion, this paper reflects on the need for collaboration among engineers, policy experts, and legal scholars to develop a comprehensive Blueprint for ADS safety and highlights opportunities for future research.