Abstract

AI systems represent a double-edged sword, capable of both empowering and harming individuals, particularly those in vulnerable positions. This chapter, based on vulnerability theory, contends that a responsible state should play a central role in overseeing and balancing the advantages and challenges that AI may present for vulnerable individuals in the digital realm. It introduces an analytical framework comprising three ethical principles to guide the governance of AI systems impacting vulnerable populations: the do-no-harm principle, the beneficence principle, and the justice principle. To operationalize these principles, the chapter puts forth a series of recommendations. In terms of the do-no-harm principle, it was suggested that the government should clearly define “significant harm” and prohibit any AI systems that exploit vulnerabilities and cause substantial harm to vulnerable individuals and advocates for the establishment of a reporting mechanism by public regulators and the adoption of an internal monitoring mechanism by private firms. Regarding the beneficence principle, the chapter proposes that the government should introduce measures to incentivize private firms to develop “orphan AI programmes”, recognize their contributions to enhancing ESG performance, and, when necessary, utilize public resources to address the lack of such programmes, thereby nurturing an inclusive and equitable technological landscape. Lastly, concerning the justice principle, the chapter urges designers and operators of AI programmes to prioritize three core values: equality and fairness, freedom of informed choice, and human flourishing, over profit maximization to achieve AI justice for vulnerable individuals. The chapter emphasizes the significance of transparency in the input datasets used for machine learning and algorithm models and calls for the establishment of a coordinated and collaborative regulatory framework by various public agencies.