Desafíos actuales de la Inteligencia Artificial

C apítulo 19 A REVIEW OF HIGH-RISK ARTIFICIAL INTELLIGENCE (AI) SYSTEMS THAT ASSESS SOCIAL SECURITY ELIGIBILITY Mariah B rochado Chair of Philosophy of Technology and Digital Rights (UFMG). Visiting Professor at the Leibniz-Institut für Medienforschung | Hans-Bredow-Institut. President of the Artificial Intelligence in Law of Minas Gerais’ Bar commission. Lucas P orto PhD in Law (UFMG). Senior researcher at the Chair of Philosophy of Technology and Digital Rights (Philotech – UFMG). Socialenvironmental Scientist. Lawyer. Amanda M apa PhD student in Law (UFMG). Master of Laws (UFOP). Research director at Brazil’s Social Security Studies Institute. Lawyer. ABSTRACT: From 2020, as debate on AI regulation by legislative bodies looms, social security agencies in France, the UK and Brazil have increasingly used automated and AI de- cision systems to assess eligibility for social security benefits. By 2024, AI legislation will take root, requiring a review of the regulatory legitimacy of federal agencies and their ongoing use of AI systems. The European Union’s AI law bans general-purpose scoring systems and classifies systems used by public authorities to assess social security benefits as high-risk, af- fecting fundamental social rights. However, current practices in AI eligibility assessment have damaged citizens’ social rights. For example, France’s CAF reported that 100,000 citizens have been wrongly denied benefits since 2021, Brazil’s INSS found an increase in benefit de- nials with automated systems, and the UK’s AI system overestimated income, leading to inap- propriate benefit denials. These practices illustrate the potential injustices that AI can cause and highlight the need for comprehensive regulatory frameworks. Cases from Australia and Brazil highlight the serious consequences of automated social protection systems, and show the need for fairness, transparency and human dignity in AI systems used in social protection. KEYWORDS: Artificial Intelligence; Social Security; Automated Decision Systems; Regulatory Framework; Eligibility Assessment.