AI Policy and Legislative Research Institute, Republic of Korea

A successful first Global Dialogue on AI Governance, convened under the United Nations, should move beyond high-level principles and deliver practical, implementable outcomes that can guide both national and international action. First, it should produce a shared baseline framework for AI governance that identifies minimum common elements across jurisdictions—such as risk classification, accountability mechanisms, and transparency standards—while allowing flexibility for local adaptation. Rather than duplicating existing principles, the Dialogue should clarify how these can be operationalized in real policy environments. Second, the Dialogue should establish a permanent multi-stakeholder coordination mechanism, building on models such as national and regional Internet Governance Forums. This would ensure continuity beyond a one-time event and enable ongoing collaboration among governments, industry, academia, and civil society. Third, a key outcome should be the development of practical toolkits for implementation, particularly for countries and local governments with limited capacity. These could include guidelines for public-sector AI procurement, regulatory sandboxes, and oversight structures, helping bridge the gap between global discussions and on-the-ground execution. Fourth, the Dialogue should promote trust-based international cooperation, including voluntary information-sharing on AI risks, incidents, and best practices. This would contribute to reducing fragmentation and fostering interoperability between governance regimes. Finally, success would mean setting a clear forward agenda, including measurable goals and thematic priorities for subsequent meetings. By combining inclusiveness with concrete outputs, the first Dialogue can lay the foundation for a credible and action-oriented global AI governance ecosystem.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

These four areas are deeply interconnected and reflect the most urgent governance gaps requiring coordinated global action. First, safe, secure and trustworthy AI is foundational. Without baseline safety and risk management standards, public trust cannot be sustained, and the benefits of AI will remain unevenly distributed. Immediate efforts are needed to operationalize safety through risk classification systems, testing protocols, and incident response mechanisms. Second, transparency, accountability, and human oversight are critical to ensuring that AI systems remain aligned with public interest. This includes explainability standards, clear assignment of responsibility across the AI lifecycle, and institutional mechanisms that enable meaningful human control, particularly in high-risk applications. Third, the protection and promotion of human rights must be embedded as a non-negotiable principle in AI governance. As AI systems increasingly affect decision-making in areas such as employment, public services, and security, safeguards against discrimination, surveillance abuse, and exclusion are essential. Finally, the broader social, economic, ethical, cultural, linguistic, and technical implications of AI must be addressed to avoid deepening global inequalities. In particular, there is an urgent need to ensure inclusiveness-especially for underrepresented languages, regions, and communities-and to support capacity-building for developing countries and local governments. Taken together, these priorities emphasize a shift from abstract principles to practical, implementable governance frameworks that are inclusive, accountable, and globally interoperable.

First, implementation capacity and governance inequality remain under-addressed. Many countries and local governments lack the institutional, technical, and financial capacity to operationalize AI governance frameworks. Without targeted support, global standards risk widening the gap between high-capacity and low-capacity actors. Second, public sector AI procurement and deployment governance is an emerging priority. Governments are rapidly adopting AI systems, yet clear standards for vendor accountability, long-term risk management, and trust-based procurement remain insufficiently developed. This is a critical entry point for embedding responsible AI in practice. Third, multi-level governance coordination-across global, national, and local levels-deserves greater focus. Current discussions are often state-centric, but many real-world AI impacts occur at the municipal or sectoral level. Mechanisms that connect these layers are essential for coherent and effective governance. Fourth, AI ecosystem concentration and market power should be addressed more directly. The dominance of a small number of technology actors raises concerns about dependency, fairness, and global inclusiveness, particularly for developing countries. Finally, continuous, multi-stakeholder governance infrastructure is needed. Rather than one-off dialogues, there is a need for institutionalized platforms that enable ongoing participation, knowledge-sharing, and iterative policy development. Addressing these cross-cutting issues will be key to translating principles into equitable, practical, and sustainable AI governance outcomes.

In the Republic of Korea, rapid AI adoption across public and private sectors is outpacing the development of practical, enforceable governance frameworks, creating both significant challenges and opportunities. A key challenge is the gap between high-level principles and real-world implementation. While Korea has advanced national strategies and strong digital infrastructure, there is still limited standardization in areas such as AI risk classification, auditing, and accountability—particularly at the level of local governments and public institutions. This leads to uneven governance capacity and potential risks in public service delivery. Another pressing issue is transparency and accountability in deployed systems. As AI is increasingly used in administrative decision-making, concerns are growing around explainability, responsibility allocation, and effective human oversight. Without clear institutional mechanisms, public trust may be weakened. At the same time, data concentration and platform dependency present structural challenges. A small number of dominant technology actors shape much of the AI ecosystem, raising concerns about fairness, innovation barriers for smaller actors, and long-term strategic autonomy. However, these challenges also create important opportunities. Korea is well-positioned to develop scalable governance models, particularly through its experience in multi-stakeholder platforms such as the Korea Internet Governance Forum. These models can support continuous dialogue, policy experimentation, and inclusive participation. Additionally, there is strong potential to lead in public-sector AI governance innovation, including trust-based procurement systems, regulatory sandboxes, and localized governance frameworks. By focusing on implementation-oriented approaches, Korea can contribute practical, globally relevant models that bridge the gap between principles and execution in AI governance.

The Global Dialogue on AI Governance, convened under the United Nations, can play a pivotal role as a practical coordination platform that bridges fragmented global efforts and moves beyond principle-setting toward implementation. First, it can function as a convergence mechanism by aligning diverse national and regional approaches to AI governance. Rather than imposing uniform standards, the Dialogue can promote interoperability through shared baseline frameworks, common terminology, and mutual understanding of risk classifications and regulatory approaches. Second, it can enable continuous multi-stakeholder engagement. By institutionalizing participation from governments, industry, academia, and civil society, the Dialogue can ensure that governance remains adaptive, inclusive, and grounded in real-world developments. Third, the Dialogue can serve as a knowledge-sharing and capacity-building hub, particularly for countries and regions with limited resources. This includes disseminating best practices, policy toolkits, and lessons learned from implementation, helping reduce global disparities in AI governance readiness. Fourth, it can foster trust-based international cooperation, including voluntary information-sharing on AI risks, incidents, and regulatory experiences. This is essential to addressing cross-border challenges and preventing regulatory fragmentation. Finally, the Dialogue can help define a forward-looking global agenda, identifying priority areas for action and facilitating coordination across existing initiatives. Its added value lies in its ability to connect discussions with implementation, ensuring that international cooperation translates into tangible governance outcomes.

The AI Dialogue should build upon and connect with existing global and multi-stakeholder initiatives to avoid duplication and maximize impact. Key frameworks include the OECD AI Principles, which provide widely recognized normative guidance, and the UNESCO Recommendation on the Ethics of Artificial Intelligence, which emphasizes human rights and inclusiveness. In addition, the Global Partnership on Artificial Intelligence (GPAI) offers practical policy insights and expert-driven research. The Dialogue should also engage with governance processes such as the Internet Governance Forum (IGF), including national and regional IGFs, which provide established models for inclusive, bottom-up participation. Furthermore, regional regulatory developments—such as the **European Union AI Act—offer valuable lessons on operationalizing risk-based approaches. The added value of the AI Dialogue lies in its ability to act as a connecting and translating platform. It can bridge high-level principles and real-world implementation by synthesizing insights from these initiatives into actionable guidance. Unlike existing frameworks that often operate in silos, the Dialogue can facilitate cross-framework interoperability, helping align standards and reduce fragmentation. Moreover, it can elevate underrepresented perspectives, particularly from developing countries and local governance actors, ensuring more equitable participation. By linking existing efforts with practical implementation tools and sustained multi-stakeholder engagement, the AI Dialogue can enhance coherence, inclusiveness, and effectiveness in global AI governance.

To ensure meaningful participation, the AI Dialogue under the United Nations should adopt a structured multi-stakeholder model where each group has a clearly defined role. Governments should contribute policy experiences, regulatory approaches, and implementation challenges. Industry should provide technical insights, risk management practices, and deployment realities. Academia can offer evidence-based research and evaluation frameworks, while civil society should represent public interest concerns, including human rights and social impacts. In terms of format, the Dialogue should move beyond plenary discussions and include thematic working groups focused on concrete issues such as AI safety, public-sector deployment, and accountability mechanisms. Each group should aim to produce practical outputs, such as policy toolkits or draft frameworks. A hybrid participation structure is essential to ensure global inclusiveness, combining in-person sessions with robust online engagement. In addition, national and regional consultations—linked to platforms like the Internet Governance Forum—can feed into the global Dialogue, creating a bottom-up process. Finally, the Dialogue should establish a continuous engagement mechanism, rather than a one-time event, enabling iterative feedback, follow-up actions, and sustained collaboration among stakeholders.

Despite progress, several key groups remain underrepresented in global AI governance discussions. First, local governments and public-sector practitioners are often excluded, even though they are responsible for implementing AI in areas such as public services, procurement, and administration. Their inclusion is essential for bridging the gap between policy and practice. Second, developing countries and smaller economies face structural barriers to participation, including limited resources and technical capacity. Without their input, global governance risks reinforcing existing inequalities. Third, linguistic and cultural minorities are insufficiently represented, particularly in discussions on data, language models, and digital inclusion. This raises concerns about bias, exclusion, and cultural homogenization. Fourth, youth and future stakeholders—who will be most affected by AI—are often engaged only symbolically rather than substantively. To address these gaps, the Dialogue should provide financial and logistical support for participation, enable remote and multilingual engagement, and create dedicated tracks or advisory groups for underrepresented communities. Partnering with national and regional platforms, such as local IGFs, can further expand inclusive participation.

To foster dynamic and meaningful engagement, the AI Dialogue should adopt innovative, outcome-oriented formats. One effective approach is policy labs or co-creation workshops, where diverse stakeholders collaboratively design solutions to specific governance challenges, such as AI procurement standards or risk assessment models. These sessions should produce tangible outputs within a limited timeframe. Another format is scenario-based simulations, where participants respond to realistic AI governance situations—such as cross-border incidents or high-risk system failures. This helps identify gaps in current frameworks and encourages practical problem-solving. The Dialogue could also incorporate "implementation clinics", where countries or organizations present real policy challenges and receive targeted feedback from experts and peers. This would directly support capacity-building and knowledge exchange. Additionally, multi-level dialogue sessions—connecting global, national, and local actors—can ensure that governance discussions reflect real-world complexity and implementation needs. Finally, digital tools can enable continuous engagement, such as open consultation platforms, collaborative drafting spaces, and real-time polling during sessions. By combining these formats, the AI Dialogue can move beyond discussion toward active collaboration, experimentation, and practical impact in global AI governance.

Several emerging policies and practices demonstrate how AI governance can move from principles to practical implementation. First, risk-baseEuropean Union AI Act-offer a structured model by categorizing AI systems according to their level of risk and applying proportionate obligations. This approach provides clarity for both regulators and developers while maintaining flexibility for innovation. Second, multi-stakeholder governance platforms, includiInternet Governance Forum (IG and national IGFs, demonstrate effective models for in Third, p are gaining importance. For e Fourth, regulatory sandboxes provide a practical tool for Fifth, AI procurement standards are an increasingly important lever. Moving from cost-based procurement to trust-based models-which emphasize long-term accountability, performance monitoring, and vendor r Finally, capacity-building initiatives-including to Together, these practices highlight the importance of combining regulatory clarity, inclusive governance, and implementation-focused tools to achieve effective