Artificial Intelligence and the Protection of Fundamental Rights
The report examines the impact of artificial intelligence systems on fundamental rights where AI is used to make decisions about people, from the provision of services and lending to candidate selection, migration and law-enforcement procedures. Its starting premise is that such an assessment has practical value only if it can change the decision to deploy a system, prevent harm and restore the position of the individual concerned. The object of analysis is therefore not the model but the decision taken with its help.
The study reviews the legal framework as of 23 October 2026: the AI Act timeline as amended by Regulation (EU) 2026/1744, including the postponed deadlines for high-risk systems; the fundamental rights impact assessment under Article 27 and its relationship with the DPIA; and the independent requirements of data protection, equality and administrative procedure. It also analyses the Council of Europe Framework Convention on AI, the HUDERIA methodology, the UN Guiding Principles on Business and Human Rights, the NIST AI RMF, the French mechanism for access to the grounds of algorithmic administrative decisions, and comparative benchmarks from the United Kingdom and the United States. The applied part covers the detection of discrimination, error metrics, meaningful human oversight and complaints, as well as police cooperation, migration, banking and sanctions screening, employment and social services.
The central conclusion is that four elements must be linked: a specific right, the mechanism of a possible violation, verifiable evidence and a protective measure. A model’s technical accuracy does not answer whether an interference is permissible, a management system certificate does not establish the lawfulness of an individual decision, and the presence of a staff member in the process does not prove that oversight is meaningful. The postponement of part of the high-risk regime does not suspend existing rights or justify delaying preparation.
Observatoire ARGA proposes a limited, evidence-based model of independent oversight comprising a research module, the review of specific situations where a mandate and lawful access exist, and the preparation of proposals for consultations, with separate legal and technical peer review. A practical programme sets out phased implementation, from a training pilot to limited practice and periodic review. The proposals for international consultations concern the minimum content of an impact assessment, the portability of explanations, the testing of human oversight, safe analysis of group disparities and the correction of decisions already taken.
The report explicitly states the limits of this model: it does not confer supervisory powers on the Observatoire or replace the decision of a competent authority; the practical scenarios and calculations are illustrative; and the proposals do not confirm submissions made, agreements with public authorities or field audits carried out.
Citation Rules
- Mandatory source attributionWhen using ARGA Observatory materials, the full name must be cited.
- Date and version indicationFor analytical reports, the year of publication is mandatory.
- Link to the originalElectronic materials must be accompanied by an active link to the official website.
- Context preservationCitations must not be shortened or altered in a way that distorts the original meaning.
- Note on adaptationIf the text is abridged or translated, state: "adapted from ARGA Observatory report".
- No commercial use without written permission from the organization
- Data accuracy preservationCharts and tables must be reproduced without changes.
Ask the ARGA experts
Four short steps. We reply within one business day.