The EU model of AI governance: regulating artificial intelligence through law and policy

ABSTRACTAll models of legal governance and most regulatory options have to do with ‘top-down’ solutions as an essential ingredient of the approach. Such models may include ‘bottom-up’ forms of self-regulation, such as in forms of ex post regulation, or unenforced self-regulation. This paper focuses on what lies in between such top-down and bottom-up approaches, namely, the middle-out interface of the analysis. Within the EU legal framework, this middle-out layer is mainly associated with forms of co-regulation, as defined by Recital 44 of the 2010 AVMS Directive and Article 5(2) of the GDPR. However, there are also additional models on how we should grasp the middle-out layer of legal regulation, as shown by the debates on the governance of AI and the Web of Data. For example, the debates on issues such as monitored self-regulation, coordination mechanisms for good AI governance, and ‘wind-rose’ models for the Web of Data make it clear that co-regulation is not the only alternative to both bottom-up and top-down approaches. From a methodological viewpoint, the middle-out approach sheds light on three different kinds of issues that regard (i) how to strike a balance between multiple regulatory systems; (ii) how to align primary and secondary rules of the law; and (iii) how to properly coordinate bottom-up and top-down policy choices. The increasing complexity of technological regulation recommends new models of governance that revolve around this middle-out analytical ground.

With the breakthrough development of artificial intelligence technology, the application of large models in government governance has shown a profound impact. This paper systematically explores the multi-dimensional application scenarios of large artificial intelligence models in constructing China's digital and intelligent government and the resulting governance transformation. The study finds that through application scenarios such as automated government affairs processing, intelligent public services, urban risk prediction, and policy simulation evaluation, large models have significantly improved the administrative efficiency and decision-making of the government and promoted the transformation of the governance model towards a data-driven flat structure. However, risks such as algorithmic discrimination, privacy leakage, administrative ethics violations, and technological dependence have emerged during the technology application process. This paper argues that the deep integration of large artificial intelligence models and government governance needs to strike a balance between technological innovation and institutional regulation. By establishing an adaptive governance framework, we can not only give full play to the advantages of technological empowerment but also avoid potential risks and ultimately achieve the modern upgrade of the national governance system and governance capabilities.

The current global cyber governance model is dominated by Western liberal norms and multi-stakeholder values. Dissatisfied with the status quo, some developing countries like China embrace another governance concept called cyber sovereignty, which advocates more state control. Meanwhile, AI development further enlarges cyberspace’s national security threats, but an international governance framework is absent in the AI realm and China is eager to take the lead in building one. This gives rise to the question: what explains China’s approach to cyber and AI governance? Current studies on cyber sovereignty and China’s AI governance model are mostly qualitative and/or have a small sampling frame, while the meaning of cyber sovereignty is debatable. Therefore, this article applies topic modelling to official/semi-official texts about cyber and AI governance to understand the cyber sovereignty concept and how it shapes China’s approach to AI governance. This article finds that cyber sovereignty is an extension of China’s state-centric view of international order. Not being a passive recipient of norms, China hopes to shape alternative cyber norms to defend national security. Now, since the global community has not reached a consensus over global AI governance, China is exploiting this gap to promote its own set of cyber visions.

The introduction of Artificial General Intelligence (AGI) brings new challenges to governance mechanisms, which ought to be able to strike a balance between innovation, moral mandates, and safety of a society. The paper discusses the importance of strategic foresight in the process of creating effective guardrails to AGI systems using contemporary literature on issues in AI governance, anticipatory governance regimes and complex adaptive system theory. We address these issues by conducting an analytical review of their regulation to date, and innovative governance models that have and are being developed, and propose an existent regulatory paradigm with the use of combined strategic forward-looking approaches to governance, especially AGI. The results indicate that conventional governance mechanisms do not match the dynamic, emergent AGI systems, requiring a changed regime of governance based on dynamic monitoring and stakeholder involvement, and anticipatory risk management. A multi-tiered governance model with technical, organizational and policy-level guardrails is presented, using empirical evidence of the existing AI governance experiences across multiple industries. The paper is a valuable addition to the body of literature on responsible AI, since it provides practical guides to anticipatory AGI governance that can adapt to changes in technology in embracing ethical standards and retaining a healthy dose of public trust.

This literature review explores the transformative potential of Artificial Intelligence (AI) in advancing sustainable development, highlighting its applications across sectors such as finance, construction, healthcare, and cultural heritage. AI’s capabilities in data processing, automation, and decision-making enable resource optimization and support progress toward the Sustainable Development Goals (SDGs). However, a major concern is the “principles-to-practices gap,” wherein high-level ethical AI frameworks lack clear implementation mechanisms, especially in low-resource or marginalized contexts. The review synthesizes global case studies, including AI deployment in mountain communities and cultural institutions, to demonstrate the value of context-sensitive, human-centric design. These examples reveal how AI can bridge digital divides and empower underrepresented groups when developed inclusively. However, risks of “AI neo-colonialism” persist, as governance models from high-income countries may marginalize diverse development needs. The review identifies shared themes such as data centrality, ethical design, and alignment with SDGs, while highlighting disparities in resources, governance models, and goals across organizations. It underscores the need for adaptive, inclusive AI governance frameworks that balance innovation with accountability. Policy implications include the need for enforceable, risk-based AI regulations, international cooperation for harmonized standards, and investment in explainable AI and infrastructure sustainability. Future research should prioritize empirical studies on governance practices, particularly in the Global South, and develop sector-specific tools to map AI’s contributions to sustainability. Ultimately, responsible AI governance must integrate social, cultural, and political dimensions to ensure that AI supports not just innovation, but equitable, inclusive, and sustainable global development. Keywords: Artificial Intelligence, Sustainable Development, Ethical Governance, SDGs

This study examines the sovereignty-internationalism paradox in AI governance, which encapsulates the tension between state control over algorithmic systems and the necessity of transnational collaboration to regulate borderless technologies. AI systems simultaneously reinforce national sovereignty while demanding global cooperation, posing a fundamental challenge to governance frameworks. Through a comparative analysis of the European Union’s AI Act and China’s Social Credit System, this research explores how AI governance can either enable state authority or necessitate transnational coordination. The EU’s risk-based regulatory model demonstrates a supranational approach balancing innovation with ethical oversight, while China’s centralized AI-driven governance exemplifies state-centric control through algorithmic surveillance. To address these tensions, the study introduces Digital Federalism, a governance model that integrates subsidiarity and multi-tiered sovereignty, enabling regulatory coordination across local, national, and global levels. Unlike polycentric governance, which emphasizes decentralization, Digital Federalism retains hierarchical coordination while allowing for adaptive, context-sensitive governance. By bridging classical sovereignty theories with contemporary digital governance debates, this paper argues that reconciling national sovereignty and transnational cooperation requires adaptive frameworks prioritizing transparency, inclusivity, and accountability. The findings illuminate the interplay between state power and global governance, providing insights for policymakers navigating AI governance frameworks in an interconnected world.

Toward responsible AI governance: Balancing multi-stakeholder perspectives on AI in healthcare.

The 2nd half of the 1990s has been a pivot point for starting a period of reform of the European higher education sector which is still continuing. It resulted in far reaching reforms, such as the introduction of the Bachelor-Master structure, but also in a revolutionary change of the learning paradigm from expert driven to student-centred education. As a result, the quality of higher education programmes has significantly improved. However, the process has also led to a lot of red tape showing a serious decline of trust in academia and its academics. This downturn can be related to the introduction of the concept of ‘new governance’ and as a follow-up the 21st century ‘modern university’ governance model in particular in Anglo-Saxon countries, but also in the Netherlands. As a consequence, the philosophy ‘profit-for-people’ has taken hold in society, including universities. This has resulted in the administrative culture crisis, which is now widely acknowledged, and has also influenced higher education institutions. Was the transfer of an industrial society to a knowledge-based economy (digital revolution) an important cause for the 1990s pivot point, the fourth industrial revolution (Artificial Intelligence) is expected to have a substantial impact on the role of the university again. This next pivot point will change this role in society fundamentally. It will also impact its governance model again, forced to give more credit to its professionals to allow for strengthening ‘academic freedom’, which has suffered as an effect of the reforms, and to open up higher education also to those who are already active in the workplace. This implies enhancing flexibility of educational programmes, the integration of the relative new phenomenon of micro-credentials (topics of specialized learning to upgrade knowledge and skills) and better alignment between higher education and society, including the world of work. It will also ask for further international cooperation at the level of academics, in parallel to management and administration, by not only covering research but also education. Past experience has demonstrated that substantial change cannot be enforced, it requires a high level of involvement and commitment of the academics involved. Academic staff is expected to keep up with the high speed of societal change which implies developing high level skills and competences and integrating the UN Sustainability Goals in their learning. A serious challenge, which has to be tackled in order to maintain relevance for society.

The impact of Internet financial regulatory policy development on green financial policy, corporate governance and corporate environment is the focus of social and economic development. It is necessary to look for more advanced risk prediction technology, and AI (Artificial Intelligence) technology has a good performance in risk analysis. It is an excellent and novel idea to combine traditional financial risk prediction technology with AI technology and use AI to optimize financial risk prediction technology. This paper proposed a financial risk prediction technology based on AI, which combined traditional financial risk prediction technology with AI technology. The algorithm proposed in this paper is a financial risk analysis algorithm based on AI. This algorithm can use AI to analyze and process a large number of financial transactions and financial risk information. This algorithm can improve the accuracy and speed of risk prediction, and can also speed up the frequency of risk prediction. According to the survey, from 2018 to 2020, with the development of Internet financial regulatory policies, the proportion of the company’s principal-agent governance model decreased from 32.08% to 26.09%, and finally to 21.13%. In addition, the proportion of the equity incentive governance model and the separation of two rights governance model had increased. Through the survey results, the changes and impacts of Internet financial regulatory policy development on green financial policies, corporate governance and enterprise environment can be clearly understood. This study can also prove the role of AI technology in financial risk prediction. In general, this research provided valuable information for the study of Internet financial regulatory policies.

Meta-Regulation: An ideal alternative to the primary responsibility as the regulatory model of generative AI in China

Background: The implementation of artificial intelligence (AI) in the public sector represents a crucial milestone in the digital transformation of Ecuador, particularly in Guayaquil. While AI has the potential to optimize administrative processes, enhance public services, and improve data-driven decision-making, its deployment without a robust governance framework poses ethical, legal, and social risks. Methods: This study employs a quantitative, non-experimental research design to analyze AI governance in Guayaquil’s public administration. Data collection was conducted through structured surveys administered to public employees, evaluating key dimensions such as governance components, public perceptions, technological infrastructure, and ethical considerations. Statistical techniques, including Pearson correlation analysis, were used to assess relationships between variables. Results: Findings highlight the disparity in AI adoption among public institutions, with significant gaps in training, infrastructure, and policy implementation. The study underscores the necessity of an AI governance model that ensures transparency, inclusivity, and ethical compliance. Conclusions: The proposed governance model provides strategic recommendations for AI adoption in Guayaquil’s public sector, emphasizing regulatory frameworks, capacity-building initiatives, and cross-sector collaboration. This research contributes to the global discourse on responsible AI governance, aligning with international efforts to establish ethical standards for emerging technologies

In the context of today’s fast-reading development of artificial intelligence (AI), the governance model supported by AI technology has become the focus of global attention. This study focuses on the AI governance model based on ethical norms in the construction of the rule of law, with the aim of providing research support for effective AI governance. In this paper, the ethical core and mechanism of AI governance in the soil of the rule of law are studied under the premise of the rule of law, and a risk governance identification model based on the DEMATEL-ISM model is constructed to analyze the risk factors of AI governance. Then, the ethical norms of AI governance are analyzed in depth, and the path of improving the ethical norms of AI governance is explored through the fsQCA method, and the AI ethical norms guidelines are constructed under the basic guideline of adhering to human-centeredness. Finally, the implementation of AI governance ethical norms is carried out, and under AI governance, the average correct rate of event classification and processing accuracy is above 85%, 207 resolved events are added every month, and the cumulative number of resolved events is increased by 2,486. At the same time, the risk factor risk can be seen, proving that the governance model under the participation of AI can effectively improve the efficiency of governance to promote the development and application of AI governance model provides a certain theoretical basis and reference value.

Health care leaders face a strategic dilemma: traditional expert-led content development ensures safety but is too slow for digital innovation, whereas artificial intelligence (AI) automation offers speed but introduces risks from hallucinations. Resolving this tension requires governance frameworks that balance operational efficiency with rigorous accountability for patient safety. This study describes the development process and conducts a formative evaluation of the Semiautomatic Clinical Algorithm Development (S-ACAD) framework as an industry-driven implementation strategy. We aimed to assess the feasibility of this "human-in-the-loop" governance model in balancing the need for operational efficiency with the rigorous safety standards required for pediatric emergency guidance. We conducted a prospective, single-day proof-of-concept case study focusing on pediatric febrile seizures. A single physician expert executed a 4-phase workflow: (1) parallel data collection using multiple AI agents, (2) AI-assisted synthesis, (3) iterative refinement via "AI sparring," and (4) final clinical validation. The resulting algorithm was reviewed by 2 independent external pediatric specialists. We benchmarked this process against a fully automated system (Fully Autonomous Clinical Algorithm Development [F-ACAD]) to illustrate comparative efficiency and safety trade-offs. In this single execution, the S-ACAD framework produced a parent-actionable febrile seizure algorithm in approximately 245 minutes. Two independent pediatric specialists (N=2) reviewed the output and did not identify medically inaccurate sections or critical safety errors requiring mandatory correction, and both rated overall clinical validity highly (9.0 and 9.5 out of 10). During the workflow, 19 human expert interventions were recorded, with clinical judgment (n=8, 42.1%) and safety review (n=5, 26.3%) as the most frequent categories in an exploratory post hoc analysis. By comparison, the fully automated approach (F-ACAD) completed the task in approximately 68 minutes, but its own AI critics identified 17 issues (9 high-priority), including concerns related to emergency response clarity and standard-of-care alignment. These preliminary findings suggest that the S-ACAD framework may offer a potential pathway for "active governance" in AI-assisted clinical content development. In this proof-of-concept case, the framework combined rapid AI-assisted drafting with continuous expert oversight and independent clinical review, suggesting the potential to reduce turnaround time while maintaining safety safeguards. However, these results are based on a single expert applying the workflow to a single clinical topic, and validation across multiple experts, topics, and institutional contexts is needed before generalizability can be established.

The increasing complexity and speed of digital transformation have challenged traditional governance models in enterprise software-as-a-service (SaaS) environments. Simultaneously, the proliferation of no-code development and the adoption of artificial intelligence (AI) across business processes have created both new opportunities and governance risks. This review presents a comprehensive theoretical framework for AI-driven agile governance—a model that integrates autonomous AI agents with no-code platforms to enable scalable, adaptive, and continuously compliant enterprise operations. The paper outlines the architecture, input features, and training methodologies of the proposed system, demonstrating how it surpasses traditional rule-based and manual governance models in accuracy, responsiveness, and auditability. Drawing from case studies, industry implementations, and comparative evaluations, we show how AI can augment governance by automating compliance enforcement, optimizing decision-making, and empowering citizen developers through secure and intelligent orchestration. The review also offers targeted recommendations for practitioners, CTOs, and policymakers, while identifying future research directions in human-AI collaboration, governance benchmarking, and cross-domain scalability. Our findings suggest that the convergence of AI and no-code platforms, under an agile governance paradigm, represents a fundamental shift in how enterprises can innovate responsibly and govern intelligently at scale.

Across the European Union, considerable discrepancies can be observed regarding the current state of AI adoption in the public sector and the complexity of functioning AI governance structures. This can be attributed to diverse levels of digitalisation, AI maturity and governance styles across EU member states. In the field of AI implementation and AI governance models in the public sector the frontrunner is the Netherlands, scoring first in the Global Index on Responsible AI. Analysing this example of good practices in terms of AI governance, with a focus on the delegation acceptance perspective, is of relevance for the state of art on AI governance within the EU. The article looks into the structure of the public Dutch Algorithm Register which currently contains over 400 entries, the AI framework for the public sector, supervisory structures in place and risks management approaches, addressing the importance of values in the development and deployment of AI systems and algorithms. The article demonstrates how in the case of AI also, early adaptors shape future behaviours, thus carrying a burden of responsibility when developing and deploying key enabling technologies in line with the core values.