INP PAN

The first Global Dialogue on AI Governance would be successful if it moves beyond broad principles and produces a credible foundation for sustained, inclusive, and action-oriented cooperation. In my view, success would mean four things. First, the Dialogue should create a shared baseline of understanding on core governance objectives: safety, human rights, accountability, transparency, and equitable access to AI benefits. This does not require full regulatory harmonisation, but it does require enough common ground to reduce fragmentation and policy gaps. Second, it should give meaningful voice to a wide range of stakeholders, especially developing countries, technical experts, civil society, academia, and smaller market actors. A successful Dialogue must not become a forum dominated only by major powers or large technology companies. Third, it should identify practical areas for cooperation, such as capacity-building, interoperability of governance approaches, technical standards, evaluation methods, incident reporting, and support for trustworthy public-interest AI. Concrete follow-up mechanisms are essential. Fourth, the Dialogue should establish a roadmap for continuity: regular engagement, thematic working streams, and measurable outputs that can inform future UN processes and national or regional initiatives. Ultimately, success would not be defined by a single declaration, but by whether the Dialogue helps build trust, reduces governance asymmetries, and creates realistic pathways for responsible AI development and use across jurisdictions. The most valuable outcome would be a process that is principled, globally representative, and capable of turning high-level commitments into practical cooperation.

• Safe, secure and trustworthy AI
• Transparency, accountability, and human oversight
• Interoperability of governance approaches
• Protection and promotion of human rights

I selected these four themes because they form the core conditions for legitimate and effective AI governance. Safe, secure and trustworthy AI is a foundational priority because governance must address not only innovation, but also technical robustness, cybersecurity, misuse prevention, and resilience across the AI lifecycle. Protection and promotion of human rights is equally essential. AI governance must remain anchored in human dignity, equality, privacy, non-discrimination, due process, and access to remedy. Without a rights-based approach, governance risks becoming purely procedural or market-driven. Transparency, accountability, and human oversight are critical because they make governance operational. It is not enough to state that AI should be trustworthy; organisations must be able to explain system purpose, allocate responsibility, document decision-making, and ensure meaningful human review where appropriate. Interoperability of governance approaches is a practical global necessity. Different jurisdictions will continue to develop distinct legal and policy frameworks, but greater compatibility between approaches can reduce fragmentation, support compliance, and enable cooperation without forcing uniformity. Together, these themes balance values and implementation. They help ensure that AI governance is not only principled, but also workable across legal systems, sectors, and levels of institutional maturity.

Yes. Several important cross-cutting issues deserve more explicit attention. First, AI governance and data governance should be linked more clearly. Questions of data quality, provenance, lawful access, data rights, and cross-border data use are central to trustworthy AI but are often treated separately. Second, compute, infrastructure, and concentration of power should be recognised as governance issues. Access to compute, cloud infrastructure, and advanced models is unevenly distributed, which may deepen global inequalities and limit meaningful participation in AI development. Third, environmental sustainability should be addressed more directly. The energy, water, and material costs of training and deploying AI systems have governance implications, particularly as adoption scales globally. Fourth, stronger attention should be paid to governance of general-purpose and open-weight models, including downstream accountability, model adaptation, and the allocation of responsibilities across the AI value chain. Finally, institutional capacity for implementation and enforcement is an emerging priority. Many jurisdictions may endorse AI principles, but lack regulators, auditors, technical expertise, or evaluation tools needed to operationalise them. These issues cut across safety, rights, accountability, and development. Addressing them explicitly would make the Dialogue more realistic, future-oriented, and responsive to the structural conditions shaping global AI governance.

Governance gaps in the areas of safe and trustworthy AI, human rights, transparency and accountability, and interoperability are already affecting countries such as Nigeria by creating uncertainty at the point of implementation. Nigeria has taken important steps through its National Artificial Intelligence Strategy and the broader digital governance agenda, while the Nigeria Data Protection Act 2023 provides an important legal foundation for rights-based data governance. However, gaps remain between high-level policy ambition and practical implementation capacity. One major challenge is fragmentation. As different governance frameworks emerge globally, countries and sectors operating across borders face growing complexity in aligning local priorities with international expectations on risk management, accountability, and safeguards. A second challenge is institutional and operational capacity: many organisations and public bodies still need stronger technical expertise, assurance mechanisms, and practical tools for oversight, documentation, and evaluation. This is closely linked to AI literacy, which is now a critical governance issue. Without sufficient AI literacy among policymakers, regulators, deployers, and users, even well-designed rules may be difficult to implement effectively. UNESCO has also highlighted AI capacity-building needs in Nigeria's public sector. At the same time, these developments create significant opportunities. In Nigeria, current efforts around AI strategy, digital transformation, and innovation support can help build a governance model that is rights-based, inclusive, and adapted to local needs, while remaining interoperable with global approaches. This creates an opportunity to invest early in AI literacy, trusted governance, and public-interest capacity so that AI adoption supports innovation, trust, and more equitable participation in the digital economy.

The AI Dialogue can play a valuable role as a global bridge between principles, policy, and practical cooperation. Its greatest contribution would be to create an inclusive multilateral space where states, international organisations, technical experts, civil society, academia, and the private sector can identify shared governance objectives while acknowledging different levels of capacity, legal traditions, and development priorities. First, the Dialogue can help build a common vocabulary around key concepts such as safety, accountability, human oversight, transparency, and human rights. This is important because international cooperation becomes much harder when jurisdictions use similar terms in different ways. Second, it can promote interoperability across governance approaches. The goal should not necessarily be full harmonisation, but greater compatibility between legal, regulatory, technical, and standards-based approaches so that fragmentation does not undermine either innovation or protection. Third, the Dialogue can support more equitable participation in AI governance by highlighting the needs of developing countries, including capacity-building, institutional readiness, AI literacy, access to technical expertise, and infrastructure constraints. In this way, it can help reduce governance asymmetries between countries that shape AI and those that are primarily affected by it. Fourth, it can encourage practical cooperation through follow-up mechanisms, such as expert working groups, shared good practices, policy toolkits, risk assessment approaches, and voluntary frameworks for transparency and assurance. Ultimately, the Dialogue should help move international cooperation from abstract consensus to usable governance pathways. Its value lies not only in convening actors, but in creating trust, shared understanding, and practical foundations for responsible, inclusive, and globally relevant AI governance.

The AI Dialogue should build on, and connect with, existing initiatives rather than duplicate them. Important foundations include the Global Digital Compact and the UN process that established both the Global Dialogue on AI Governance and the Independent International Scientific Panel on AI; together, these provide the clearest multilateral anchor for ongoing cooperation. It should also connect with the UNESCO Recommendation on the Ethics of AI and related UNESCO policy dialogues, which have already advanced globally recognised normative principles and implementation support. In addition, the Dialogue should draw on the OECD AI Principles, the OECD's analytical and measurement work, and the OECD-UN collaboration announced to strengthen global AI governance cooperation. Relevant technical and policy communities such as ITU's AI governance work should also be linked, especially on operational issues such as standards, capacity-building, and governance tools. The added value of the AI Dialogue should be its ability to connect these efforts within one inclusive UN platform. Unlike smaller or more like-minded forums, it can bring together states from all regions alongside civil society, academia, technical experts, and the private sector in a setting with broader legitimacy and representation. Its distinctive contribution should be threefold: first, to promote interoperability across existing governance initiatives; second, to identify gaps where current mechanisms remain fragmented or inaccessible, especially for developing countries; and third, to translate existing principles into practical cooperation, including capacity-building, shared terminology, policy toolkits, and governance pathways that are globally relevant but sensitive to different national contexts

Different stakeholders should contribute in ways that reflect their distinct roles and expertise. States can share regulatory experience, policy priorities, and implementation challenges. International organisations can help connect the Dialogue with existing global frameworks and capacity-building efforts. Technical experts and academia can contribute evidence, risk analysis, evaluation methods, and foresight. Civil society can bring rights-based perspectives, lived experience, and accountability concerns. The private sector can share operational lessons from development, deployment, assurance, and governance practice. To be effective, the AI Dialogue should combine plenary visibility with smaller, outcome-oriented working formats. A useful structure would include high-level plenary sessions, thematic breakout discussions, and cross-cutting working groups focused on practical outputs. Each thematic track should include a balanced mix of governments, experts, civil society, and industry. Written submissions should feed into discussions in a transparent way. The Dialogue should also be designed as a continuing process, not a one-off event. It would benefit from clear documentation, rapporteur summaries, and follow-up mechanisms such as expert groups, issue papers, and capacity-building streams. Regional consultations before and after the main Dialogue would also help ensure broader participation and relevance. Overall, the structure should be inclusive, multidisciplinary, and geared toward practical cooperation rather than only general statements of principle.

Several voices remain underrepresented in global AI governance discussions. These include stakeholders from the Global South, smaller and lower-capacity states, local regulators, public-interest technologists, workers and trade unions, educators, children and young people, persons with disabilities, minority language communities, and communities most affected by automated decision-making without having a direct seat at the table. There is also a persistent gap between high-level global discussion and the perspectives of those responsible for implementation, such as public administrations, SMEs, auditors, procurement teams, and sector-specific practitioners in health, education, labour, and justice. Their experience is essential because many governance challenges emerge in practice rather than at the level of abstract principles. Inclusion requires more than symbolic representation. It means funding participation, supporting multilingual access, enabling remote engagement, publishing accessible background materials, and creating structured opportunities for input before decisions are shaped. Regional and sectoral consultations should feed directly into the Dialogue. Selection processes for speakers and contributors should be transparent and designed to ensure geographic, disciplinary, gender, and linguistic diversity. Targeted inclusion of affected communities and practitioners would make the Dialogue more legitimate, grounded, and responsive to real-world governance needs.

The most effective formats will be those that move beyond formal statements and enable real exchange across disciplines and regions. One useful model would be moderated problem-solving sessions built around concrete governance dilemmas, such as foundation model oversight, cross-border enforcement, AI literacy, or public-sector procurement. These are often more productive than abstract panel discussions. Another valuable format would be multi-stakeholder roundtables with balanced representation from governments, civil society, academia, technical experts, and industry, organised around specific questions and expected outputs. Scenario-based exercises and case clinics could also help participants test how different governance approaches work in practice, especially in high-impact sectors. To widen participation, the Dialogue could include hybrid regional hubs, interactive digital consultations, and short evidence sessions where affected communities, youth representatives, or practitioners present frontline experience. Lightning interventions, curated expert response panels, and facilitated workshops could make the process more dynamic while preserving substance. A particularly useful innovation would be to combine deliberation with drafting: small mixed groups could work on non-binding issue notes, terminology maps, or practical recommendations during the event itself. This would help turn discussion into tangible outputs. The most meaningful format is therefore one that is participatory, multilingual, problem-oriented, and designed to produce usable follow-up rather than only dialogue for its own sake.

One useful example is the Artificial Intelligence Impact Assessment (AIIA) highlighted by the OECD AI Policy Observatory. It helps organisations translate responsible AI principles into operational controls by structuring risk identification, accountability, and mitigation in a practical assessment format. This kind of tool is valuable because it turns abstract governance commitments into concrete decision-making processes. A second important example is the Council of Europe's HUDERIA methodology, which offers a structured way to assess AI systems from the perspective of human rights, democracy, and the rule of law. Its added value lies in connecting technical and organisational risk management with broader societal impacts, while also aiming for compatibility with other frameworks and standards. This makes it a strong model for interoperable and rights-based AI governance. A third example comes from the European Research Council (ERC), which has recently clarified limits on the use of AI in grant evaluation and broader research processes. The ERC approach is important because it shows how sector-specific governance can preserve human responsibility, integrity, and trust while still allowing carefully bounded use of AI tools. This is a useful model for governance in science, education, and other high-trust domains. Taken together, these examples show that effective AI governance is most credible when it is practical, rights-aware, context-sensitive, and enforceable in real organisational settings. The AI Dialogue could add value by promoting exchange between such initiatives, identifying shared good practices, and encouraging adaptation of these models across jurisdictions and sectors.