Safeguarding human dignity: A narrative review of prohibited AI practices under the EU AI Act

Artificial intelligence (AI) technologies are reshaping public administration, law enforcement, and social governance—but not without raising profound human rights concerns. This narrative review examines the eight AI practices explicitly prohibited under Article 5 of the European Union Artificial Intelligence Act (EU AI Act), which categorizes them as presenting an “unacceptable risk.” These practices include subliminal manipulation, exploitation of vulnerable populations, social scoring, predictive profiling, untargeted scraping of biometric data, emotion recognition in sensitive settings, biometric categorization by sensitive attributes, and real-time biometric surveillance in public spaces. The purpose of this review is to assess how each prohibition corresponds to specific human rights protections, such as autonomy, privacy, non-discrimination, and dignity, and to explore the legal and ethical frameworks that justify such prohibitions. This study employs a qualitative narrative methodology, integrating legal analysis, historical misuse cases, and ethical theory—drawing from sources including the EU Charter of Fundamental Rights, the European Convention on Human Rights, and scholarly work in AI ethics. Key findings reveal that each prohibited AI practice has precedent in past abuses and can be normatively justified using deontological, utilitarian, and virtue ethics frameworks. The review concludes that the prohibited AI systems not only breach legal standards but undermine the moral foundations of democratic societies. These findings support the necessity of rights-based AI regulation and underscore the EU’s global leadership in normative governance. Future research should focus on enforcement challenges, international harmonization, and the development of new safeguards for emerging AI risks.