AI Governance in a Complex and Rapidly Changing Regulatory Landscape: A Global Perspective

AI presents unique regulatory challenges that require flexible and adaptive governance as nations race to develop the technology amid uncertainty. Its rapid growth and increasingly consequential decision-making roles drive a push for governance grounded in human oversight, understanding, and ethics to build trust. Laws are needed due to unpredictability and identified risks, such as manipulation of behavior, exploitation of vulnerabilities, and surveillance. However, rules must avoid stifling innovation. While national rules matter, a global framework is better suited to transnational limitations and could provide model laws for harmonization. The global AI landscape risks increasing disparities between developed and developing countries, raising debates over AI as a state-controlled resource versus common heritage. Competing interests—innovation and growth versus precaution—complicate international regulation. The authors propose that a new global AI regulatory authority could anticipate and respond to technological change, balance stakeholder interests, provide technical expertise, and help share benefits while minimizing risks. The paper outlines a theoretical framework, reviews literature, analyzes roles of key actors (states, individuals, civil society, companies, international organizations), details how a regulatory authority could establish and monitor international law, and maps the road ahead.

Debate over AI regulation has persisted for decades, with a growing race among nations and organizations to shape governance. Scholars disagree on whether and how to regulate, the sufficiency of existing legal tools, and the role of international law versus domestic approaches. Key themes include: definitional uncertainty; competing regulatory theories (public interest, organizational, capitalist state); and AI’s cross-sector reach affecting numerous regimes (trade, human rights, IP, humanitarian law, nuclear security, private international law). Some argue AI will reshape or stress the limits of international law, while others see opportunities to use international law to guide AI development and enforcement. Proposals range from soft-law principles and ethics frameworks to binding treaties and new institutions. Approaches include risk-based regulation, regulatory sandboxes, sector-specific versus general rules, and soft versus hard law. Scholars also debate regulatory efficiency, timing of intervention, and hybrid public–private governance. The literature identifies challenges of transparency, accountability, and enforcement, and notes fragmentation across multiple institutions and norms. There is growing attention to smart regulation, precautionary approaches, and ensuring that governance addresses societal values and global inequalities. The paper also surveys major initiatives and regulations: the EU AI Act (2024), which sets harmonized rules, risk-based classification, governance structures, and enforcement mechanisms; the Council of Europe’s Framework Convention on AI, Human Rights, Democracy and Rule of Law (2024), proposing binding obligations and oversight; UNESCO’s (2021) Recommendation on the Ethics of AI outlining values, principles, and policy actions; China’s regulatory measures (e.g., Deep Synthesis Provisions, algorithmic recommendation rules, and interim measures on generative AI); and the United States’ federal and state actions (e.g., Executive Order, AI Bill of Rights). It highlights multilateral efforts (OECD, ISO, IEEE, UN bodies) and civil society and research initiatives shaping norms, while noting limitations such as lack of Global South representation and compliance incentives.

Conceptual and theoretical framework development. The authors construct a governance framework centered on a prospective global AI regulatory authority and its interactions with key actors (states, international organizations, corporations, and civil society). Drawing on international law theory and historical experience from other transnational issues (climate change, transboundary resources), the framework examines: rulemaking (hard and soft law), coordination and standard-setting, and monitoring/enforcement support. It considers structural features of international law—state consent, sovereignty, legitimacy, and compliance dynamics—and maps institutional design questions (independence, accountability, political control, quality of regulation). A schematic (Fig. 1) positions the authority within a network of stakeholders performing lawmaking, monitoring, coordination, ethical oversight, innovation/compliance, and public engagement. This theoretical approach is complemented by a narrative review of legal scholarship and existing initiatives to assess feasible pathways toward global AI governance.

- International regulation of AI is necessary given cross-border risks (security, accountability, bias, privacy, cyber threats) and benefits, and to avoid fragmented, conflicting national approaches. - A new global AI regulatory authority could coordinate standards, facilitate dialogue, promulgate soft-law instruments initially, and support domestic implementation and monitoring of international commitments. - Soft-law instruments (recommendations, guidelines, standards) are pragmatic first steps due to rapid technological change, sovereignty concerns, and geopolitical competition; nonetheless, binding instruments are emerging regionally (EU AI Act 2024; Council of Europe Framework Convention 2024). - Effective governance requires clarifying roles and responsibilities of multiple actors: states as primary law-makers and implementers; companies as developers/operators needing oversight and incentives; civil society as ethical watchdogs and contributors to agenda setting; and international organizations as conveners and norm entrepreneurs. - Monitoring and enforcement at national level remain challenging; international mechanisms should provide guidance, capacity-building, peer review, and transparency, mindful of North–South capacity gaps. - Centralized versus decentralized architectures each have trade-offs; proposals range from a single coordinating body to a networked, multi-institutional approach. - Existing initiatives (UNGA Res. 78/265; UNESCO Recommendation; OECD principles; regional acts) can be leveraged as models toward future global harmonization. - Major obstacles include divergent national interests, race dynamics, legitimacy and representation concerns (particularly for lower-income countries), and uncertainties regarding legal personality, liability, and AI personhood.

The study argues that a coherent transnational framework, spearheaded by a global AI regulatory authority, best addresses AI’s cross-border risks and opportunities. By emphasizing soft law in the near term, the framework balances innovation with protection while building consensus toward more binding norms. The proposed authority would catalyze standard-setting, coordinate fragmented initiatives, and provide technical and legal support for national implementation and monitoring—addressing compliance gaps rooted in state consent, capacity, and legitimacy. Mapping stakeholder roles shows how shared responsibilities can mitigate regulatory inertia and forum fragmentation, while incorporating ethical oversight and public engagement to bolster trust. Leveraging existing regional and multilateral instruments can accelerate harmonization and reduce the risks of regulatory competition. Overall, the approach connects international law’s strengths (norm diffusion, coordination, reputational incentives) to AI’s governance needs, acknowledging practical constraints and geopolitical realities.

The article presents a theoretical framework for global AI governance, contending that complex, rapidly evolving risks require an internationally coordinated response and likely a new regulatory authority. It underscores that progress should start with soft-law tools and existing initiatives (e.g., EU AI Act, Council of Europe Framework Convention, UNESCO, OECD), moving gradually toward binding global norms as consensus develops. Recommendations include: fostering interdisciplinary collaboration among legal, engineering, and computer science experts; expanding communication and participation among all stakeholders; using EU and Council of Europe instruments as models for future treaties; and conducting further studies on the feasibility, design, and impact of a global AI regulatory authority. The authors conclude that while the debate is nascent and obstacles remain, international law can provide needed clarity, legitimacy, and harmonization for AI governance.

The study is primarily conceptual and focuses on global developments, without systematically analyzing the interplay between international and national laws. It emphasizes legal perspectives over political and economic analyses and relies largely on Western-centric instruments, with limited representation from the Global South (apart from the Chinese example). The evolving international landscape and emerging actors may alter the proposed framework. The absence of new empirical data limits long-term assessments. These constraints may affect generalizability and completeness of policy implications.