A survey of artificial intelligence risk assessment methodologies

Abstract

Artificial Intelligence (AI) systems are increasingly being deployed in a wide range of applications. For many of these, a failure of the system would be embarrassing and inconvenient, but otherwise harmless. Failures in other applications, such as credit assessments, recruitment, medical diagnostics, (semi-)autonomous vehicles and power grid load distribution systems, however, would have significant impacts on the lives of individuals, organizations or large parts of society.