A Structured Approach to Identifying and Characterizing AI Vulnerabilities - RAND

A research organization is developing a framework to identify and understand vulnerabilities in AI systems. This matters for public safety and security as AI systems become more widely used in critical applications.