The systematic process of identifying, analyzing, evaluating, and mitigating the potential adverse outcomes and vulnerabilities associated with AI systems. This includes addressing risks related to security, bias, transparency, and operational failures.