AI-Driven Public Services: A Taxonomy of Accountability and Sovereign Artificial Intelligence (AI)

Abstract

This study examines the institutional and technological challenges of integrating artificial intelligence (AI) systems into public administration and governmental services, focusing on the taxonomy of algorithmic roles in decision-making, the balance of interests in cooperation with commercial AI providers and infrastructure actors, and the safeguarding of national technological sovereignty. A qualitative interdisciplinary approach is applied, combining regulatory and legal analysis, thematic examination of empirical cases across different countries, and theoretical synthesis. Data were collected from official documents, peer-reviewed publications, and news sources, using snowball sampling for case selection and iterative coding for analytical categorization. The research develops a six-tier pyramidal model of accountability distribution according to the degree of algorithmic autonomy in decision-making chains: from full delegation («AI as Captain»), provision of ready-made solutions for human approval («AI as Navigator»), configuration of option sets («AI as Adviser»), environmental analysis with trigger signaling («AI as Observer»), execution of labor-intensive tasks under operator supervision («AI as Workforce»), to routine operational support without decision-making capacity («AI as Routine Assistant»). The model is mapped against risk gradations (high, limited, minimal) to assess error consequences.The findings reveal the dilemma of public-private partnerships, which facilitate access to innovation but simultaneously reinforce dependence and systemic vulnerabilities. The study also substantiates the role of sovereign AI as a strategic response to these risks. For effective integration of AI into governmental services, it recommends mandatory classification of systems by autonomy and criticality levels. The proposed six-level taxonomy enables a differentiated approach to accountability allocation, reducing institutional gaps and risks of bias, while enhancing resilience and strategic security.