A structured process for evaluating the potential effects of AI systems on individuals, groups, organisations, and society before development or deployment. Impact assessments identify and address potential risks, harms, and unintended consequences.

AI impact assessments provide a proactive approach to identifying and mitigating potential negative effects of AI systems. Comprehensive assessments consider diverse impact dimensions, including privacy implications, fairness and bias concerns, security risks, environmental impacts, and potential job displacement. By conducting assessments early in the development process, organisations can make design choices that minimise harms and maximise benefits before significant resources are invested or systems are deployed.

‍

A city government conducting an impact assessment before deploying an AI-powered public transportation optimisation system, examining how route changes might affect access for low-income communities, potential privacy implications of mobility data collection, and environmental impacts of changing transportation patterns.