Mercer County Community College

A successful first Global Dialogue on AI Governance would do more than restate high-level principles. It would clarify where meaningful international convergence is possible now and identify concrete next steps for cooperation, capacity-building, and implementation. Three core outcomes recommendations: First, the Dialogue should sharpen shared understanding of what effective governance requires in practice. That includes moving beyond nominal commitments to "human oversight" and toward clearer expectations around competence, authority, accountability, contestability, and the institutional conditions needed for oversight to be real. Second, it should advance a more globally usable approach to capacity-building. Capacity-building should not be limited to technical access or participation alone. It should also include support for risk interpretation, supervisory competence, incident review, public-interest oversight, and governance design, especially in settings where resource asymmetries or dependence on external providers may weaken local control. Third, the Dialogue should identify practical areas for continued multilateral work. These include responsibility mapping across the socio-technical stack, context-sensitive thresholds for adequacy and deployment, and mechanisms for transparency, remedy, and review that are meaningful for affected persons and institutions. A successful first session would therefore leave participants with more than a record of discussion. It would produce a clearer common agenda: where further cooperation is needed, what governance capacities deserve priority support, and which institutional safeguards are most urgent for responsible and inclusive AI governance.

• AI capacity-building
• Interoperability of governance approaches
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

These four areas are closely connected and, taken together, address some of the most urgent gaps in current AI governance. Transparency, accountability, and human oversight is a priority because many governance frameworks still treat oversight too formally. A human may be assigned responsibility without having the competence, authority, time, or institutional support needed to exercise meaningful judgment. This remains a major practical weakness. AI capacity-building is equally urgent because effective governance depends on more than technical infrastructure. It also requires capacities for risk interpretation, supervisory review, incident response, contestation, and public-interest oversight. Without stronger institutional and human capacities, even well-designed frameworks may remain aspirational. Protection and promotion of human rights is essential because failures of accountability, opacity, and inadequate review often fall most heavily on persons and communities with the least power to challenge decisions or seek remedy. Human rights protections should remain central to international governance discussions, especially where AI systems affect public services, labor, migration, education, or civic participation. Social, economic, ethical, cultural, linguistic and technical implications of AI matters because the international landscape is already developing unevenly across jurisdictions, sectors, and regulatory traditions. Greater interoperability can support cooperation without requiring full uniformity. It can also help reduce fragmentation, improve comparability, and make it easier to share governance tools, lessons, and safeguards across different institutional settings. Together, these priorities support a practical model of AI governance that is more implementable, more inclusive, and more accountable.

Yes. One important cross-cutting issue is the difference between nominal human involvement and effective human oversight. Although human oversight appears in the listed themes, a persistent governance gap remains under-specified: institutions may formally keep humans "in the loop" while still weakening judgment through automation deference, deskilling, throughput pressure, unclear authority, or lack of escalation pathways. This is not only a technical or organizational issue. It is a question of whether institutions remain capable of responsible judgment in practice. A second cross-cutting issue is responsibility localization across the socio-technical stack. The governance object may be described as a model, application, interface, workflow, vendor service, or broader institutional arrangement. When the object of governance is under-specified, responsibility can become diffuse. Stronger international attention to responsibility mapping would help clarify where judgment is delegated, where authority appears, where intervention is possible, and where accountability should reattach when harms occur. A third emerging issue is the need for proactive augmentation of inherited oversight capacities. In many contexts, traditional administrative, analog, and earlier digital governance practices remain important foundations. At the same time, they may be over-matched by the scale, speed, opacity, and authority effects of contemporary AI systems. This suggests that governance support should be approached proactively, not only reactively after harms or dependencies have become entrenched. These issues cut across several listed themes and would strengthen the Dialogue's practical relevance.

In my country, region, and sector, and globally, the central challenge is that AI deployment is advancing faster than governance readiness. Institutions increasingly have access to powerful systems, but often lack clear standards for oversight, responsibility, review, and remedy. This creates a widening gap between technical capability and accountable use. One major effect of this gap is that "human oversight" is often treated formally rather than substantively. Responsibility may be assigned without sufficient competence, authority, time, or procedural support. In practice, this can produce over-reliance, automation deference, deskilling, and weak accountability, particularly in settings shaped by speed, scale, and institutional pressure. Another challenge is uneven governance capacity across institutions and regions. Some actors can invest in evaluation, monitoring, and governance design, while others remain more dependent on external providers, opaque systems, or inherited administrative processes not designed for contemporary AI. This can deepen asymmetries in control, scrutiny, and meaningful participation. At the same time, these pressures have created important opportunities. They are making clear that governance must include capacity-building, not only principle-setting; effective oversight, not only nominal human involvement; and clearer responsibility mapping across the socio-technical stack. They also create an opening for international cooperation around practical safeguards, institutional support, and governance approaches that preserve human judgment, accountability, and meaningful local agency.

The AI Dialogue can play a valuable coordinating role by helping states and stakeholders move from broad principle agreement toward more practical forms of cooperation. The United Nations has positioned the Dialogue as part of the wider Global Digital Compact architecture, with the first session in Geneva in July 2026 and a second session planned for New York in May 2027. That creates a useful space not only for exchange, but for building continuity across discussions that are otherwise fragmented across regions, sectors, and institutions. Its greatest value may be in clarifying where international convergence is already possible. This includes shared expectations around effective human oversight, transparency, accountability, human rights protections, and capacity-building. The Dialogue can also help identify which issues require deeper coordination across jurisdictions, including responsibility mapping across the socio-technical stack, meaningful contestation and remedy, and more comparable ways of assessing whether systems are fit for particular uses. The Dialogue can also strengthen cooperation by connecting principle-setting with implementation support. Many institutions, especially in lower-capacity settings, need more than high-level commitments. They need usable governance tools, peer learning, capacity-building support, and clearer pathways for adapting international guidance to local institutional realities. A United Nations forum is well placed to encourage this kind of bridge-building across technical, legal, policy, and public-interest communities. Its added value, then, is not to replace existing initiatives, but to connect them, widen participation, and help build a more inclusive and practically usable basis for international cooperation on AI governance.

The AI Dialogue should build on existing international initiatives that already provide important normative, technical, and institutional foundations. These include the United Nations Global Digital Compact, the UNESCO Recommendation on the Ethics of Artificial Intelligence, the OECD AI Principles, the Council of Europe Framework Convention on Artificial Intelligence and human rights, democracy and the rule of law, and the G7 Hiroshima AI Process, including its more recent reporting framework for organizational transparency. Each contributes something important: global political legitimacy, ethical guidance, practical governance principles, legal development, and more structured reporting on risk-management practices. The Dialogue should also connect with implementation-oriented mechanisms such as UNESCO's readiness and ethical impact assessment tools, as well as broader United Nations and International Telecommunication Union ecosystems that support peer learning, technical exchange, and practical cooperation around AI for sustainable development. These are useful because many governance gaps are not only normative but institutional: states and sectors often need support in translating principles into oversight practices, review processes, and governance capacity. The Dialogue's added value would be to link these initiatives more effectively across communities that do not always work together: intergovernmental, regulatory, technical, academic, civil society, and sectoral actors. It can provide a more inclusive forum for identifying overlap, reducing fragmentation, and highlighting practical governance gaps that cut across existing frameworks, especially around effective human oversight, capacity-building, accountability, and interoperability. In that sense, the Dialogue can serve less as a new standalone regime than as a connective platform that improves coherence, participation, and practical uptake across the international AI governance landscape.

Different stakeholders should contribute in differentiated but connected ways. Governments can bring regulatory experience, public-sector implementation challenges, and lessons from national or regional governance efforts. Technical communities can clarify system capabilities, limits, evaluation practices, and emerging risks. Civil society and human rights organizations can identify lived harms, accountability gaps, and barriers to remedy. Academic and research communities can contribute comparative analysis, evidence, and longer-horizon reflection. Private-sector actors can share operational experience, deployment realities, and governance practices, while being expected to do so with enough transparency to support meaningful dialogue. The United Nations has framed the process as a multi-stakeholder consultation for the Global Dialogue on AI Governance, which makes this kind of differentiated participation especially important. For format and structure, the Dialogue would benefit from a layered design rather than a single plenary-only model. A strong structure could include: (1) short plenary sessions focused on major cross-cutting questions; (2) smaller thematic working sessions organized around the priority areas; (3) structured multi-stakeholder round-tables designed to surface areas of convergence, disagreement, and implementation gaps; and (4) a synthesis session that identifies practical next steps, capacity-building needs, and topics for continued cooperation. To make the Dialogue more useful, contributions should not be limited to general statements of principle. Participants should be encouraged to share governance experiences, implementation challenges, concrete practices, and lessons learned. It would also help to provide a light common template for interventions, for example: key challenge, practical implication, existing response, remaining gap, and recommendation. That would improve comparability while still allowing flexibility across sectors and regions.

Several voices remain underrepresented in global discussions on AI governance, even where multi-stakeholder participation is formally emphasized. These include stakeholders from lower-capacity states and institutions, especially in parts of Africa, Latin America, small island developing states, and other settings where resource constraints and dependence on external providers can limit meaningful influence. Underrepresented voices also include frontline public-service practitioners, educators, labor representatives, disability advocates, linguistically marginalized communities, smaller civil society organizations, and people directly affected by AI-mediated decisions but not usually included in governance design. The wider United Nations process around the Global Dialogue has emphasized inclusion and stronger participation by developing countries, and that priority should remain central in practice. Inclusion should not be treated only as invitation. It also requires conditions for participation that are actually usable. This includes advance circulation of materials, multilingual access, remote participation options, travel support where possible, plain-language summaries, and enough time and structure for less-resourced participants to prepare substantive contributions. It also helps to create targeted sessions for perspectives that might otherwise be overshadowed in more general discussion. A second issue is that under-representation is not only geographic. Some of the most important missing perspectives concern how AI governance works in everyday institutions: schools, hospitals, welfare systems, local administration, workplaces, and civic information settings. Including these voices would improve practical relevance by grounding discussion in real oversight conditions, not only in high-level policy or technical debate.

The most effective engagement formats will likely be those that move beyond serial speeches and allow participants to work through concrete governance problems together. Because the Dialogue is meant to support multistakeholder exchange and practical cooperation, it would benefit from formats that combine structured comparison with genuine interaction. One useful format would be case-based governance labs, where diverse participants respond to a common scenario such as public-sector deployment, labor-related use, education, healthcare triage, or cross-border data and model dependence. This helps reveal where governance principles align, where they diverge, and what implementation gaps remain. A second useful format would be challenge-response roundtables. In these, one stakeholder group presents a concrete governance difficulty, and other groups respond with recommendations, cautions, and examples from their own sectors or regions. This can make engagement more dynamic and comparative. A third option would be small facilitated synthesis groups tasked with identifying three things: areas of convergence, unresolved tensions, and priority next steps for cooperation. Those outputs could then feed into a closing synthesis session. It may also be valuable to include light structured polling or ranking exercises before or during sessions, so participants can identify which governance gaps, capacity needs, or implementation barriers they see as most urgent. This would create a clearer picture of shared priorities without reducing the Dialogue to voting alone. The best innovative formats will be those that keep discussion grounded in practice, make room for underrepresented voices, and produce outputs that can inform follow-up cooperation rather than ending with general statements alone.

Several existing policies, practices, and platforms offer useful building blocks for effective AI governance. One important example is the NIST AI Risk Management Framework and its related profiles. Its value lies in giving organizations a repeatable, lifecycle-based structure for identifying, assessing, and managing AI risks, rather than treating governance as a one-time compliance exercise. The newer work around sector-specific profiles also shows how general governance principles can be translated into more use-specific guidance. A second strong example is the UNESCO Recommendation on the Ethics of Artificial Intelligence, together with UNESCO's Readiness Assessment Methodology. These are useful because they connect ethical principles with country-level readiness, institutional capacity, and practical assessment. They help move discussion from abstract commitments toward implementation conditions, especially for states and sectors still building governance capacity. A third example is the OECD/G7 Hiroshima AI Process Reporting Framework, which offers a more structured approach to transparency and accountability for organizations developing advanced AI systems. This kind of reporting mechanism is valuable because it encourages comparability across organizations and jurisdictions while still allowing for different governance models. A fourth example is the Council of Europe Framework Convention on Artificial Intelligence and human rights, democracy and the rule of law. Its importance lies in linking AI governance directly to human rights, democracy, and rule-of-law commitments through an international legal instrument. Finally, the OECD AI Incidents and Hazards Monitor is a useful practice-oriented platform because it helps build an evidence base from real incidents and hazards rather than relying only on anticipated risks. Taken together, these examples show that effective AI governance is strongest when it combines principles, implementation tools, reporting mechanisms, legal safeguards, and real-world learning.