UN Report Calls for Interoperable AI Governance Frameworks

The United Nations University Institute in Macau (UNU Macau) has released a comprehensive policy report titled Interoperability in AI Safety Governance: Ethics, Regulations, and Standards, offering a comparative analysis of AI governance frameworks across China, South Korea, Singapore, and the United Kingdom. Published on 15 October 2025 and formally launched at UNU’s 2025 AI Conference on 4 December 2025, the 122-page report examines how these four technologically advanced yet distinct jurisdictions regulate artificial intelligence in three high-risk sectors: autonomous driving, education, and cross-border data flows. At a time when global AI governance remains fragmented and regulatory divergence threatens international cooperation, the report advances interoperability as a pragmatic middle path—one that avoids both the rigidity of full harmonisation and the chaos of regulatory silos.

Why it matters. For executives, policymakers, and AI developers, the report signals a critical shift in compliance strategy. As nations implement divergent regulatory models—from the EU AI Act to China’s algorithmic governance regime—companies can no longer afford jurisdiction-specific compliance. Instead, they must design systems that function across regulatory boundaries. The report underscores that interoperability is not merely a technical challenge but a multidimensional imperative spanning ethics, regulation, and standards. With AI increasingly embedded in critical infrastructure and transnational services, the ability to align governance across borders will determine both market access and operational resilience. The report’s alignment with the Global Digital Compact (GDC), adopted by all UN member states in 2024, further elevates its relevance as a normative reference point in global digital policy.

Ethical Convergence, Regulatory Divergence

The report finds that while China, South Korea, Singapore, and the UK broadly agree on core ethical principles—such as fairness, transparency, accountability, and human oversight—these shared values rarely translate into consistent regulatory outcomes. For instance, in the domain of autonomous driving, all four jurisdictions require risk assessments and human-in-the-loop mechanisms, but differ sharply in enforcement mechanisms and liability frameworks. Singapore emphasizes sandboxed experimentation and outcome-based regulation, while China mandates pre-deployment certification and algorithmic transparency reports. The UK, post-Brexit, is developing a context-specific, sector-led approach, whereas South Korea enforces strict data localization and algorithmic impact assessments.

This ethical convergence with regulatory divergence highlights a central tension in global AI governance: the difficulty of translating abstract principles into enforceable rules. The report attributes this gap to differing legal traditions, economic priorities, and risk perceptions. Notably, the Global South remains underrepresented in standard-setting forums, limiting the inclusivity of emerging norms. The authors argue that without deliberate efforts to bridge these divides, interoperability will remain aspirational rather than operational.

Barriers to Interoperability

The report identifies three primary barriers to achieving meaningful interoperability in AI safety governance. First, fragmented regulatory regimes create compliance burdens for multinational firms. A self-driving vehicle approved in the UK may not meet China’s data sovereignty requirements or South Korea’s real-time monitoring mandates. Similarly, AI-powered educational tools deployed across borders face conflicting rules on student data privacy and algorithmic profiling.

Second, limited global coordination hampers the development of shared technical standards. While bodies like ISO/IEC JTC 1/SC 42 and IEEE are advancing AI standards, adoption remains uneven. The report notes that China often develops parallel standards through its national bodies, while Singapore and the UK more frequently align with international norms. This divergence complicates efforts to build interoperable systems, particularly in cross-border data flows, where data protection regimes vary significantly.

Third, the underrepresentation of the Global South in AI governance discussions undermines legitimacy and inclusivity. Despite the report’s focus on four Asian and one European jurisdiction, the authors stress that true interoperability must incorporate diverse socio-legal contexts and developmental needs. Without broader participation, governance frameworks risk entrenching existing power imbalances in the global AI ecosystem.

A Framework for Interoperable Governance

To address these challenges, the report proposes a three-dimensional model of interoperability: ethical, regulatory, and technical. Each dimension is accompanied by targeted recommendations aimed at fostering coherence without uniformity.

On the ethical dimension, the authors call for a global AI ethics framework under the auspices of the UN, building on the principles of the Global Digital Compact. Such a framework would not replace national regulations but provide a common normative baseline—similar to the UNESCO Recommendation on the Ethics of AI, but with greater emphasis on implementation and monitoring.

For regulatory interoperability, the report advocates for multilateral coordination mechanisms, such as a network of national AI safety authorities that share best practices, conduct joint audits, and recognize each other’s compliance assessments. Drawing inspiration from mutual recognition agreements in trade and telecommunications, the model promotes regulatory equivalence over identical rules. The report also recommends the development of interoperable digital public infrastructure—such as shared data trusts or federated learning platforms—that enable cross-border AI deployment while respecting local regulations.

On the technical side, the report emphasizes “interoperability by design”—embedding compatibility into AI systems from the outset. This includes adopting modular architectures, open APIs, and standardized metadata formats. The authors highlight the need for technical standards that support explainability, portability, and auditability across jurisdictions. For example, in education, an AI tutoring system should be able to adapt its data handling practices based on the user’s location, without requiring a complete redesign.

Reactions and Implications

The report was developed under the leadership of Yik Chan Chin from Beijing Normal University, Jingbo Huang and Serge Stinckwich from UNU Macau, with country-specific contributions from David A Raho (UK), Chunli Bi (China), Hag-Min Kim (South Korea), and James Ong (Singapore). It was sponsored by SenseTime, a leading Chinese AI company, a detail that may invite scrutiny given the firm’s involvement in surveillance technologies. However, the report maintains an academic tone and does not endorse specific commercial practices.

Stakeholders have responded cautiously. AI governance experts have welcomed the report’s pragmatic approach, particularly its rejection of one-size-fits-all solutions. “The report correctly identifies that harmonisation is neither feasible nor desirable,” said one digital policy analyst familiar with the GDC negotiations. “Interoperability offers a more realistic path forward.” However, some civil society groups have raised concerns about the lack of transparency regarding the quantitative data underpinning the analysis—public summaries do not disclose the number of regulations or standards mapped, nor the methodology for assessing alignment.

For AI developers and corporate leaders, the implications are clear: future compliance will require adaptive, modular governance strategies. Companies investing in autonomous systems, AI-driven education platforms, or cross-border data services must anticipate regulatory pluralism and design for flexibility. The report’s call for evidence-based, outcomes-oriented regulation suggests a move away from checklist compliance toward performance-based oversight—a shift already evident in the UK’s AI regulation white paper and Singapore’s Model AI Governance Framework.

Looking ahead, the success of interoperability will depend on sustained multilateral engagement and institutional innovation. The report positions the Global Digital Compact as a key vehicle for advancing these goals, but implementation remains uneven. As AI capabilities continue to outpace governance, frameworks like this one offer a roadmap for coherence in an increasingly fragmented landscape. The challenge now lies not in diagnosing the problem, but in mobilizing the political will and technical cooperation needed to build systems that are not only safe, but globally interoperable.