EU AI Act sparks worldwide explosion in AI governance research

A new study shows that the global research landscape on AI regulation and public governance expanded dramatically after 2022, driven largely by the rise of transformer-based AI systems and the emergence of major regulatory initiatives such as the European Union AI Act. The findings reveal a field increasingly centered on accountability, transparency, legal oversight, and public-sector governance as AI systems move deeper into administrative decision-making and public services.

The study, titled "AI spring and its regulation discourse: a bibliometric study of trends in literature," was published in Frontiers in Artificial Intelligence and examined 209 scientific publications from 2018 to 2025 indexed in the Web of Science database. The researchers traced publication trends, influential contributors, institutional collaborations, citation structures, and emerging themes shaping the global AI governance debate. The analysis identified Europe as the dominant intellectual center of AI governance scholarship while revealing growing concerns over fragmented regulation, weak institutional integration, and the urgent need for enforceable accountability mechanisms.

Transformer-era AI drives new governance and accountability concerns

Based on the analysis, 2018 marked the beginning of a new governance era for artificial intelligence because it coincided with the rise of transformer-based AI architectures following the release of Google's transformer model in late 2017. These systems dramatically accelerated AI capabilities across natural language processing, reinforcement learning, automated decision-making, and multimodal applications, triggering a surge in real-world deployment and public-sector integration.

This technological transition pushed scholarly debate beyond purely technical discussions into broader legal and societal concerns involving accountability, transparency, privacy, discrimination, and governance. AI systems increasingly became embedded in public administration, digital government, healthcare, and administrative law, forcing policymakers and researchers to examine how algorithmic systems interact with democratic oversight and institutional responsibility.

Publication activity remained modest between 2018 and 2021 but accelerated sharply after 2022. The number of annual publications rose from just three papers in 2018 and 2019 to 45 papers in 2024 and 86 papers in 2025. The researchers linked this expansion directly to the intensification of regulatory activity worldwide, particularly the development of the EU AI Act and OECD governance frameworks.

The research landscape is heavily concentrated in Europe and other advanced economies with active regulatory ecosystems. The United States led publication output with 49 papers, followed by the United Kingdom with 37, Italy with 36, Spain with 33, and Germany with 29. The Netherlands, Portugal, Sweden, France, and Australia also emerged as major contributors, while countries such as China, India, Ukraine, and Romania showed growing but comparatively limited participation.

Institutional collaboration networks further reinforced Europe's dominance in shaping AI governance debates. Universities such as University College London, the University of Oxford, the University of Pisa, and the University of Freiburg emerged as central hubs in global collaboration networks. The researchers described the system as a "European-centered research ecosystem" increasingly connected to institutions in the United States and Australia.

The analysis also revealed that foundational research published between 2018 and 2021 continues to dominate citation structures. Early works by scholars such as Luciano Floridi, Margot Kaminski, and Patrick Henman established the conceptual framework for later debates on accountability, transparency, ethical governance, and legal compliance. These earlier studies remain among the most heavily cited publications in the field, indicating that the first wave of AI governance scholarship still shapes current policy discussions.

Legal accountability and transparency dominate AI governance research

The bibliometric mapping showed that AI governance scholarship has become strongly anchored around legal and regulatory concepts rather than purely technical innovation. The most frequent keywords across the literature included "Artificial Intelligence," "Law," "Transparency," "Governance," "Management," "Data Protection," and "Accountability."

The study identified accountability, human rights, and societal impact as the field's most developed "motor themes," meaning they are both highly influential and conceptually cohesive within the research network. In contrast, broader concepts such as governance, technology, and law functioned as foundational "basic themes" that connect multiple subfields but remain less internally integrated.

The keyword co-occurrence analysis showed especially strong conceptual relationships between law and technology, governance and transparency, and data protection and privacy. The researchers concluded that the field increasingly views legal structures and governance frameworks as mechanisms for implementing ethical principles such as fairness, accountability, and transparency into AI deployment.

Much of the literature operationalizes AI governance primarily through legal accountability systems. This includes discussions of liability rules, compliance obligations, procedural safeguards, judicial review, auditability requirements, and enforceable oversight structures. Public-sector accountability emerged as a particularly important concern as governments adopt AI systems for administrative decision-making.

Despite this growing legal focus, the researchers found significant thematic fragmentation across the field. Science mapping identified seven major but loosely connected research clusters covering organizational accountability, data protection, algorithmic fairness, institutional legitimacy, public-sector risk management, autonomous systems, and geopolitical governance.

Rather than converging into a unified discipline, these subfields continue to evolve in parallel with relatively weak conceptual integration. The researchers warned that this fragmentation limits cumulative theory-building and creates gaps between legal theory, technical system design, and public administration practice.

The analysis of article abstracts reinforced this conclusion. Frequently recurring terms included "legal," "governance," "regulation," "European," "transparency," "risk," "privacy," and "human rights," showing that scholarship increasingly frames AI governance through a legal-regulatory lens tied closely to European policy development.

The European Union's legislative activity, especially around the AI Act and GDPR, has become a dominant reference point in global governance debates. The prominence of terms such as "EU," "European," and "Act" in the abstract analysis suggests that the European model now functions as one of the central organizing frameworks for AI governance research worldwide.

Fragmented governance landscape exposes institutional and policy gaps

Although AI governance scholarship has expanded rapidly, major institutional and conceptual gaps remain unresolved. One of the most significant weaknesses identified by the researchers is the limited integration between legal scholarship, public administration research, and technical AI development.

The field remains dominated by conceptual and policy-oriented work, while empirical studies examining how AI accountability systems operate in real administrative environments remain relatively scarce. The researchers found that legal and policy analyses dominate nearly all thematic areas, while practical examinations of institutional implementation, organizational capacity, and operational governance mechanisms are far less developed.

Themes such as legitimacy, institutional complexity, procedural safeguards, and rights protections for automated decision-making systems were identified as especially underdeveloped despite their growing importance for democratic governance. The researchers warned that AI deployment in public administration is not simply a technical efficiency issue but also a question of maintaining public trust, democratic legitimacy, and citizen oversight.

The study highlighted concerns that algorithmic systems may create mismatches between established administrative norms and the logic embedded within AI systems. These tensions could generate governance failures if institutions adopt AI technologies faster than they can build appropriate oversight structures.

The researchers also emphasized that the global AI governance landscape remains geographically imbalanced. Europe and North America continue to dominate research output and institutional leadership, while contributions from many regions remain limited. This imbalance raises concerns that future AI governance systems may reflect the priorities and regulatory assumptions of a relatively narrow set of political and institutional contexts.

Another major gap identified involves emerging technologies such as multimodal AI systems and foundation models. Although these technologies increasingly influence high-stakes decision-making, they remain only marginally represented in current governance research clusters. The researchers called for future studies examining how these advanced systems interact with administrative law, accountability frameworks, transparency obligations, and risk classification schemes.

As stated in the research, future progress will depend on bridging disciplinary divides and developing governance models that connect legal accountability systems with practical administrative implementation. Policymakers, institutions, and researchers must strengthen oversight capacities, improve interdisciplinary collaboration, and ensure that accountability mechanisms evolve alongside increasingly capable AI technologies.