East Asia Technology Institute

The first global dialogue on AI governance would be successful if it produces a practical, practical coordination agenda rather than only a high-level statement of principles. 1. It should clarify how global AI governance can become more interoperable. AI rules are emerging quickly, but they are fragmented across national laws, sectoral regulations, standards bodies, and voluntary frameworks. A successful dialogue should identify where common terminology, risk categories, transparency expectations, incident reporting, and accountability mechanisms can be better aligned without forcing every country into one model. 2. It should treat digital inclusion as core AI governance infrastructure in East Asia. Korea's Digital Inclusion Act shows that AI governance cannot stop at model safety or innovation policy. People encounter AI through public services, kiosks, apps, workplaces, schools, hospitals, etc. If elderly users, disabled persons, rural communities, migrants, small businesses, and digitally vulnerable groups cannot understand, access, contest, or benefit from these systems, AI governance has failed in practise. 3. The dialogue should recognise the need for regional translation layers. East Asia is central to the global AI ecosystem through semiconductors, advanced digital infrastructure, frontier AI adoption, and major technology firms. Japan, Korea, Taiwan, China, and ASEAN adjacent economies have different governance transitions and policy tools. A successful UN process should support regional observatories or coordination platforms that compare frameworks, monitor implementation, and bring regional evidence into global discussions. 4. Finally, success should mean stronger participation from underrepresented jurisdictions and communities. The dialogue should help move AI governance from fragmented principles towards measurable implementation, shared learning, and practical accountability. In short, the first global dialogue should establish the UN as a convening layer of interoperable, inclusive, and regionally grounded AI governance.

Social, economic, ethical, cultural, linguistic and technical implications of AI;Interoperability of governance approaches;Transparency, accountability, and human oversight;AI capacity-building

EATI selected these priorities because AI governance must address both technical risk and the wider social conditions that determine whether societies can benefit from AI fairly. First, the social, economic, ethical, cultural, linguistic, and technical implications of AI are urgent. AI is likely to widen existing gaps between people, firms, and countries that have access to advanced AI tools, data, compute, and skills, and those that do not. In East Asia, this issue also has a strong linguistic and cultural dimension. Korean, Chinese, Japanese, and other regional languages are not only communication tools; they carry distinct writing systems, social expressions, institutional concepts, cultural references, and historical memory. As large AI models are trained on multilingual data, governance should ensure that linguistic and cultural data are not treated simply as extractable resources. Communities and countries should have clearer rights and mechanisms around consent, licensing, representation, benefit-sharing, and digital sovereignty. Second, interoperability of governance approaches is essential. AI rules are developing quickly across national laws, sectoral regulations, standards, and voluntary frameworks. East Asia is strategically central to AI through semiconductors, digital infrastructure, major technology firms, and high AI adoption, but its governance approaches remain fragmented. Better interoperability would support cross-border accountability without forcing every jurisdiction into a single model. Third, transparency, accountability, and human oversight are necessary for trustworthy AI. Institutions need documentation, auditability, explainability, incident reporting, contestability, and clear responsibility when AI systems affect people's rights, opportunities, or access to services. Finally, AI capacity building is critical. Many governments, SMEs, civil society actors, and vulnerable communities lack the institutional and technical capacity to participate meaningfully in AI governance. Global dialogue should therefore support regional observatories, shared learning, and practical implementation capacity, especially in underrepresented regions.

First, linguistic and cultural data sovereignty should be treated as an emerging AI governance issue. In East Asia, languages such as Korean, Chinese, Japanese, and other regional languages are not only communication tools. They contain distinct writing systems, honorifics, social registers, idioms, institutional concepts, historical memory, and culturally specific reasoning patterns. If these languages are used to train AI systems without clear rules on consent, licensing, provenance, representation, and benefit-sharing, countries and communities risk losing control over core cultural infrastructure. Initiatives such as African language data projects show that language datasets can be designed as tools of digital sovereignty, not merely as raw material for global AI firms. East Asia needs comparable governance conversations around Korean, Chinese, Japanese, Taiwanese, minority, and regional language data. Second, AI governance literacy among startups, SMEs, and entrepreneurs is under-addressed. Much AI innovation is not produced only by large technology companies or governments. It is also produced by founders, product teams, small businesses, and applied-AI startups that often move faster than regulation and may not understand national, regional, or international governance frameworks until they face a compliance barrier. This creates risks around data protection, bias, cybersecurity, transparency, human oversight, and cross-border deployment. The Global Dialogue should therefore address not only state-level and corporate governance, but also the practical governance capacity of the innovation ecosystem. A useful outcome would be shared guidance, regional observatories, and implementation tools that help smaller AI actors understand their responsibilities before harm or regulatory conflict occurs.

East Asia is affected by these governance gaps in a particularly concentrated way. The region is central to AI through semiconductors, advanced digital infrastructure, major technology firms, high AI adoption, and rapidly developing national AI policies. However, its governance approaches remain fragmented. Korea is moving toward a more formal AI governance model through its AI Basic Act and Digital Inclusion Act; Japan has taken a more innovation-oriented and soft-law approach; Taiwan is building AI governance around innovation, sovereignty, and security; and China has developed more state-directed rules for generative AI and data governance. This creates a major challenge for cross-border accountability, regional interoperability, and smaller companies trying to understand their obligations. The most significant challenge is implementation. Many startups, SMEs, public institutions, and civil society actors do not yet have the technical, legal, or institutional capacity to translate AI principles into real practice. This can lead to weak documentation, unclear human oversight, poor incident reporting, inadequate accessibility, and limited understanding of how AI systems affect vulnerable communities. Another challenge is linguistic and cultural data governance. East Asian languages such as Korean, Chinese, Japanese, Taiwanese languages, and minority languages contain distinct writing systems, social registers, institutional concepts, and cultural context. These should not be treated merely as raw training data for global AI systems. Governance should address provenance, licensing, representation, benefit-sharing, and digital sovereignty. The opportunity is that East Asia can become a practical testbed for interoperable and inclusive AI governance. Because the region combines advanced technology capacity with diverse regulatory models, it can generate lessons for the wider world. Regional observatories, shared implementation tools, multilingual governance research, and startup-facing guidance could help bridge global principles and local practice.

The AI Dialogue can play a central role as a global coordination layer for AI governance. It can help governments, regional bodies, civil society, academia, industry, and technical communities identify where cooperation is most urgent and where governance frameworks can become more interoperable. First, the Dialogue can create a shared reference point for terminology, risk categories, transparency expectations, incident reporting, accountability, and human oversight. This would help reduce fragmentation while still respecting different legal systems and governance traditions. Second, it can connect global principles with regional implementation. In East Asia, countries such as Korea, Japan, Taiwan, and China have different AI governance approaches, but they are deeply connected through semiconductors, supply chains, digital infrastructure, trade, and AI deployment. The Dialogue can support regional observatories and coordination platforms that compare frameworks, track implementation, and bring regional evidence into global discussions. Third, the Dialogue can strengthen capacity building. Many governments, startups, SMEs, public institutions, and civil society actors do not yet have the technical or institutional capacity to govern AI effectively. International cooperation should therefore include practical guidance, shared tools, training, and implementation support. Fourth, the Dialogue can help make AI governance more inclusive. It should bring underrepresented countries, linguistic communities, vulnerable groups, and smaller innovators into discussions that are often dominated by large states and major technology companies. Overall, the AI Dialogue can advance international cooperation by turning fragmented AI principles into shared learning, interoperable governance, and practical implementation capacity across regions.

The AI Dialogue should build on existing intergovernmental, research, civil-society, and technical mechanisms rather than duplicate them. Relevant partners include OECD/GPAI and OECD.AI, UNESCO, ITU's AI for Good platform, the International Network of AI Safety Institutes, ASEAN's AI governance work, the Council of Europe's AI Convention, and expert institutions such as the Future of Life Institute, CSIS, CSET, Stanford HAI, and the Ada Lovelace Institute. The added value of the AI Dialogue would be connective: mapping overlaps and gaps, reducing duplication, supporting interoperability, and ensuring that smaller countries, civil society, startups, SMEs, educators, and linguistic communities are not excluded from agenda-setting. One important gap is regional translation. Global observatories exist, and Southeast Asia has emerging AI governance mechanisms, but there appears to be a need for a more dedicated, neutral, multilingual East Asia-focused governance platform covering Korea, Japan, Taiwan, China, and adjacent economies. East Asia is central to AI through semiconductors, digital infrastructure, advanced adoption, and major technology firms, yet its governance debates are often fragmented by language, legal system, and geopolitical context. Emerging institutions such as the East Asia Technology Institute are attempting to fill this gap by developing comparative, regionally grounded AI governance analysis. The AI Dialogue could add value by creating partnership and funding pathways for such regional institutions, allowing local evidence to inform global governance. The Dialogue should also connect AI governance with human agency, critical thinking, data sovereignty, and personal data control. AI literacy should help people question outputs, understand limitations, protect personal data, and preserve independent judgment as AI becomes embedded in education, work, public services, and daily life

The Dialogue should be structured so participation is meaningful rather than symbolic. Member States should identify national priorities, regulatory gaps, and areas where international cooperation is needed. International organizations and standards bodies should help map existing frameworks, terminology, safety practices, and capacity-building tools. Regional institutions should translate global principles into local implementation, compare governance approaches, and surface region-specific challenges. Academia and think tanks should provide independent evidence on risks, benefits, social impacts, and governance options. Industry, startups, and SMEs should contribute practical experience on deployment, compliance barriers, innovation constraints, and responsible product development. Civil society, educators, workers, disabled persons, youth, linguistic communities, and other affected groups should help assess how AI systems affect rights, access, human agency, critical thinking, and everyday life. The Dialogue should therefore combine plenary sessions with smaller thematic and regional working sessions. Each thematic session should include a cross-stakeholder mix rather than separating governments, industry, civil society, and technical experts into silos. Sessions should be co-chaired by a Member State and a relevant non-governmental stakeholder where appropriate. The structure should also include written submissions, multilingual participation, accessible virtual options, low-bandwidth access, travel support for under-resourced participants, and public synthesis reports explaining how stakeholder input was considered. Regional consultations before and after the Dialogue would help connect global discussions with local realities. Finally, the Dialogue should create a follow-up mechanism. This could include an implementation tracker, a public repository of governance tools, and pathways for regional observatories and smaller institutions to contribute evidence between annual meetings. This would help the Dialogue become a continuing governance process, not a one-time event.

Global AI governance discussions still underrepresent the people and institutions most exposed to AI's real-world effects. First, lower-income and smaller countries remain underrepresented, especially those without strong regulatory agencies, compute access, domestic AI industries, or permanent participation in international AI forums. Their absence matters because AI rules may still shape their economies, public services, labour markets, education systems, and data resources. Second, linguistic and cultural communities are underrepresented. AI governance is often discussed in English and by actors from a small number of technologically powerful jurisdictions. Low-resource languages, Indigenous languages, minority languages, and culturally specific knowledge systems need stronger representation. This is also relevant in East Asia, where Korean, Chinese, Japanese, Taiwanese languages, and minority languages carry distinct writing systems, social registers, historical memory, and institutional concepts. Third, digitally vulnerable groups are underrepresented: elderly people, persons with disabilities, rural communities, migrants, low-income users, children, workers, educators, and people who rely on public services. Fourth, healthcare and public-health voices need stronger inclusion, including patients, clinicians, caregivers, disabled persons, rare-disease communities, and low-resource health systems. AI systems increasingly affect diagnosis, triage, insurance, public-health surveillance, and medical research, but affected communities are not always central in governance design. Fifth, conflict-affected communities and humanitarian actors should be included. Military AI, autonomous systems, surveillance, and AI-enabled targeting raise serious questions for human oversight, civilian protection, and international humanitarian law. Finally, startups, SMEs, and local public institutions are underrepresented, despite being major AI deployers. They should be included through funded participation, multilingual submissions, accessible virtual formats, regional consultations, civil-society-led sessions, health and humanitarian tracks, and formal channels for regional observatories to submit evidence.

The AI Dialogue should move beyond traditional plenary sessions by adding structured AI Governance Scenario Labs. These would be small, facilitated breakout sessions alongside the main Dialogue, using prepared case studies and a standard reporting template. This format is realistic because UN and multistakeholder processes have already used workshops, consultations, best-practice forums, and capacity-building sessions. The innovation would be applying these formats directly to AI governance stress-testing. Each lab should bring together governments, civil society, technical experts, startups, SMEs, educators, youth, disabled persons, linguistic communities, workers, healthcare voices, and regional institutions. Participants would examine realistic AI governance scenarios, such as a public-sector AI system denying benefits unfairly, a multilingual model trained on cultural data without clear licensing, an AI health tool deployed in a low-resource setting, or a startup expanding across borders without understanding governance obligations. The purpose would not be abstract debate. Participants would identify affected groups, accountability gaps, human oversight needs, data-governance issues, accessibility failures, and possible remedies. Human rapporteurs should verify and synthesize the outputs. AI tools could assist with multilingual summarization, translation, and issue-mapping, but final outputs should remain human-reviewed. The Dialogue could also include regional deliberation rooms before and after the main event, allowing evidence from East Asia, Africa, Latin America, Europe, the Middle East, and small island states to feed into the global process. A useful final output would be a living governance-gap map showing where international cooperation is needed: interoperability, capacity-building, language-data sovereignty, startup governance literacy, human oversight, incident reporting, digital inclusion, and protection of human agency.

Effective AI governance should combine law, technical tools, institutional capacity, and regional translation. One useful policy example is Korea's Digital Inclusion Act. It shows that AI governance should not only focus on model safety, but also on whether people can access and use digital and automated services in practice. This is especially relevant for elderly users, disabled persons, rural communities, and digitally vulnerable groups. Technical risk-management tools are also important. The NIST AI Risk Management Framework offers a practical approach for identifying and managing AI risks across design, development, deployment, and evaluation. Singapore's AI Verify provides another concrete model: a testing framework and toolkit that helps organizations assess responsible AI implementation against internationally recognized governance principles. At the international level, OECD.AI and the OECD AI Incidents Monitor are useful because they turn AI governance into evidence-based monitoring, not only abstract principles. Incident tracking can help policymakers understand how harms occur, where accountability gaps emerge, and which systems require stronger oversight. Regional approaches are also necessary. The ASEAN Guide on AI Governance and Ethics is a useful example because it aims to support practical organizational guidance while encouraging alignment and interoperability across jurisdictions. Similar regional translation mechanisms are needed in East Asia, where Korea, Japan, Taiwan, China, and adjacent economies are central to AI but have different regulatory traditions and policy tools. Finally, AI governance should include education and human agency. UNESCO's AI ethics and competency work shows the importance of preparing people not only to use AI, but to question outputs, understand limitations, protect rights, and preserve critical thinking. EATI's view is that effective AI governance requires a layered approach: global principles, regional observatories, practical tools for organizations, incident monitoring, digital inclusion, and capacity-building for smaller actors.