Why AI Governance Must Start With Harms
Modern AI systems are now embedded in decision-making across finance, healthcare, employment, policing, education and online platforms, often at scale and with little human visibility into how outputs are produced. When such systems are not governed responsibly, they can produce concrete harms to individuals, groups, organizations and democratic institutions, ranging from privacy violations and discriminatory outcomes to safety failures and information disorder. The Knowledge places “identifying the types of risks and harms posed by AI to individuals, groups, organizations and society” at the foundation of AI governance, underscoring that a risk-based understanding of harms is a prerequisite for compliant and trustworthy AI programs.
Print Page