Fundamental rights and artificial intelligence impact assessment: A new quantitative methodology in the upcoming era of AI Act

Abstract

The EU Artificial Intelligence Act requires that deployers of Artificial Intelligence (AI) systems perform a Fundamental Rights Impact Assessment (FRIA) for some high-risk AI systems identified in Art. 27 of the regulation. The aim of this work is to offer a comprehensive framework to assess the impact of AI systems on Fundamental Rights (FR) of individuals. In a nutshell, the assessment approach that we propose consists of two stages: (1) an open-ended questionnaire that helps gather the contextual information and the technical features, in order to properly identify potential threats for FR, and (2) a quantitative matrix that considers each right guaranteed by the European Charter of Fundamentals Rights and tries to measure the potential impacts with a traceable and robust procedure. In light of an increasingly pervasive use of AI systems and considering the specificity of such technologies, we believe that a structured and quantitative process for assessing the impact on FR of individuals is still lacking and could be of great importance in discovering and remedying possible violations. Indeed, the proposed framework could allow to: (1) be accountable and transparent in assessing the risks of implementing AI systems that affect people; (2) gain insights to understand if any right is threatened or any group of people is more vulnerable; (3) put in place, if necessary, remediation strategies before the deployment of AI systems through demonstrable mitigative actions, with the aim of being compliant with the regulation and limiting reputational damage.