Preventing Risks to Citizens from the Use of AI by Means of Fundamental Rights Impact Assessment

ABSTRACT Artificial intelligence is here to stay; however, the implementation of these systems must be done with the utmost respect for citizens’ rights, even more so when it comes to fundamental rights. To this end, it is crucial to introduce tools in favour of these guarantees and, undoubtedly, one of them is the Fundamental Rights Impact Assessment (FRIA) report, provided for in the new European AI Regulation. This is an assessment that certain high-risk AI systems, especially those in the public sector, must undergo, the purpose of which is for the person responsible for deployment to identify the specific risks to the rights of individuals or groups of individuals who may be affected and sets out the measures to be taken in the event that these risks materialise. The role that this report is called upon to play is therefore crucial if we want a trustworthy IA. The FRIA report is discussed in the article in light of the present day development of control with AI in the member states and focuses exclusively on the public sector. Precisely the deployment of artificial intelligence (AI) in the public sector presents a series of distinctive characteristics that warrant particular consideration. Accordingly, prior to engaging in an analysis of the FRIA report, this paper provides a preliminary examination of the broader context, as well as the potential benefits and inherent risks associated with the use of AI by public administrations. This foundational analysis seeks to highlight the specific features of public sector applications that introduce significant nuances when compared to the private sector. A proper understanding of these particularities is essential to fully appreciate the relevance and implications of the FRIA report within this domain.

KEYWORDS Risks - AI - Fundamental rights - Public authorities and auditing

TABLE OF CONTENTS 1. Approach. – 2. Context. – 3. From bureaucracy to AI: the eternal balance between efficiency and safeguards. – 3.1. Logical AI or connectionist AI?. – 3.2. Regulated powers and discretionary powers. – 3.3. Better services or more control?. – 3.4. Effectiveness versus safeguards. – 4. The risks and opportunities of using AI in the public sector. – 5. The approach adopted by the European AI Regulation. – 6. The Fundamental Rights Impact Assessment report. – 6.1. Objective scope: which AI systems are required to be controlled?. – 6.2. Subjective scope: the specific nature of the public sector. – 6.3. Content of the evaluation. – 7. Conclusions.