Independent Researcher in AI Governance and Responsible AI

In my view, the first Global Dialogue on AI Governance would be successful if it achieves three outcomes. **First, it should establish a clear and credible shared understanding of what "AI governance" actually includes.** The Dialogue should avoid becoming too diffuse or drifting into broad conversations about AI in general. Its value lies in focusing specifically on governance: accountability, oversight, human control, institutional responsibility, implementation capacity, and international cooperation. **Second, it should move the conversation from principles to operational reality.** A meaningful outcome would be a stronger global recognition that the challenge is no longer the absence of AI principles, but the difficulty of implementing governance in practice. The Dialogue should surface concrete governance challenges such as institutional readiness, human oversight design, role clarity, accountability mechanisms, and the governance of AI once it becomes embedded in real decision environments. **Third, it should create a genuinely inclusive and implementation-oriented foundation for future cooperation.** This means ensuring that perspectives from emerging and regulated economies are not treated as peripheral, but as central to the discussion. AI governance cannot be globally credible if it is shaped only by a narrow set of institutional realities. The Dialogue should therefore help build a more practical and globally relevant basis for cooperation, one that reflects differences in regulatory maturity, institutional capacity, and governance readiness. In short, success would mean that the first Dialogue does not end as a symbolic discussion, but as the beginning of a more focused, operational, and internationally credible approach to AI governance.

• Transparency, accountability, and human oversight
• AI capacity-building
• Interoperability of governance approaches
• Safe, secure and trustworthy AI

I selected these priorities because, in my view, the most urgent challenge in AI governance is no longer the absence of principles, but the difficulty of translating them into operational institutional practice. **Transparency, accountability, and human oversight** are central because AI governance ultimately depends on who is responsible, how decisions are overseen, and how human judgment is preserved once AI becomes embedded in real workflows and decision environments. **AI capacity-building** is essential because many institutions - especially in regulated and emerging-economy contexts - do not yet have the governance maturity, internal capabilities, or implementation structures required to govern AI effectively in practice. **Interoperability of governance approaches** is also a key priority because AI governance cannot advance through fragmented or isolated models alone. Greater alignment across frameworks, standards, and institutional approaches will be important for building globally credible and practically usable governance ecosystems. Finally, **safe, secure and trustworthy AI** remains a core priority because governance must ultimately support systems that are not only innovative, but also dependable, resilient, and socially legitimate in real-world use. Taken together, these priorities reflect a practical view of AI governance: one that emphasizes implementation, institutional readiness, operational accountability, and globally relevant governance design.

Yes. One important cross-cutting issue that deserves greater attention is **institutional readiness for AI governance**. Many current discussions focus on principles, risks, regulation, or technical capabilities, but less attention is given to whether institutions are actually prepared to govern AI in practice. In many cases, the core challenge is not simply what should be governed, but whether organizations and public institutions have the internal structures, competencies, accountability pathways, and governance operating models required to do so effectively. This issue cuts across nearly all other themes. It affects transparency, oversight, safety, trustworthiness, interoperability, and implementation capacity. Without institutional readiness, even well-designed governance frameworks may remain largely symbolic or difficult to operationalize. A related emerging issue is the governance of AI once it becomes embedded into **decision environments**, rather than used as a standalone tool. As AI increasingly shapes workflows, judgments, escalation paths, and organizational behavior, governance must also address questions such as trust calibration, authority boundaries, role clarity, and the changing nature of human oversight in AI-mediated settings. For this reason, I believe the Dialogue would benefit from explicitly recognizing **governance readiness** and **AI in decision environments** as important cross-cutting themes for future work.

In regulated and emerging-economy contexts, particularly across parts of **Africa and the Arab region**, AI governance gaps are increasingly visible as AI adoption moves faster than institutional preparedness. One of the most significant challenges is that AI is often introduced into organizations and public-sector settings before adequate governance structures are in place. In many cases, institutions are still developing basic capacities related to data governance, risk ownership, accountability, internal oversight, procurement controls, and role clarity. As a result, AI may be adopted in ways that create uncertainty around responsibility, weaken trust, or outpace the ability of institutions to govern its use effectively. A second challenge is the uneven maturity of governance ecosystems. Different countries and sectors are moving at different speeds in terms of regulatory development, technical readiness, and institutional capability. This creates fragmentation and makes it harder to align on practical governance approaches across jurisdictions and organizations. At the same time, this moment also creates a major opportunity. Because many institutions are still in relatively early stages of AI adoption, there is a real opportunity to build governance more intentionally from the start rather than retrofitting it after harmful patterns are already embedded. This is especially important in regulated sectors and public-facing institutional environments. There is also a strong opportunity for international cooperation that is implementation-oriented — not only sharing principles, but also practical governance models, institutional design patterns, and capacity-building approaches that can be adapted to local realities.

The AI Dialogue can play a highly valuable role by serving as a **practical global coordination platform** for AI governance — one that goes beyond broad principles and supports more coherent, inclusive, and implementation-oriented international cooperation. First, it can help create a more **shared global understanding of AI governance priorities**, particularly around accountability, human oversight, institutional responsibility, governance readiness, and the conditions required for trustworthy AI deployment. Second, the Dialogue can support **cross-regional learning** by bringing together different governance experiences, including those from emerging and regulated-economy contexts that are often underrepresented in global AI discussions. This is important because meaningful cooperation depends not only on alignment, but also on understanding differences in institutional capacity, regulatory maturity, and implementation realities. Third, it can help advance **interoperability across governance approaches** by surfacing common principles, operational challenges, and practical governance patterns that can be adapted across jurisdictions without assuming identical legal or institutional models. Most importantly, the Dialogue can help shift international cooperation from a primarily declarative mode toward a more implementation-oriented one. In addition to discussion, it can become a space for sharing governance practices, institutional design approaches, oversight lessons, and practical pathways for responsible AI deployment. In that sense, the AI Dialogue can play a unique role: not by replacing existing initiatives, but by helping connect them, contextualize them, and make them more globally usable.

The AI Dialogue should build upon and connect with the substantial body of work that already exists across international, regional, and technical governance ecosystems. This includes, for example: * the **OECD AI Principles**, * **UNESCO's Recommendation on the Ethics of AI**, * the **G7 Hiroshima AI Process**, * relevant **G20** discussions, * the **NIST AI Risk Management Framework**, * and standards-related work through **ISO/IEC** and other technical governance bodies. The Dialogue should also remain attentive to evolving regional and regulatory initiatives, including approaches emerging from the **European Union** and other jurisdiction-specific governance efforts. Its added value, however, should not be to duplicate or compete with these initiatives. Rather, the AI Dialogue can add value in three important ways. **First**, it can serve as a more inclusive global space for connecting policy, governance, standards, and implementation discussions that are currently spread across different forums. **Second**, it can help translate existing frameworks into more practical and globally adaptable governance understanding, especially for countries and institutions that are still building implementation capacity. **Third**, it can elevate underrepresented perspectives — particularly from emerging and regulated-economy contexts — and help ensure that international AI governance is shaped not only by those with the most mature ecosystems, but also by those facing the most urgent implementation realities. In that sense, the Dialogue's unique contribution would be to connect, contextualize, and operationalize existing AI governance efforts in a more globally relevant and implementation-oriented way.

Different stakeholders can contribute most effectively to the AI Dialogue if participation is designed to be **structured, role-sensitive, and implementation-oriented**, rather than purely symbolic or discussion-heavy. Governments can contribute policy perspectives, regulatory experiences, and public-sector implementation challenges. Academia can provide conceptual clarity, evidence, and critical analysis. Technical communities and standards bodies can contribute practical insights on system design, safety, evaluation, and operational governance. Private sector actors can share implementation lessons and deployment realities, while civil society can help ensure that questions of legitimacy, inclusion, and accountability remain central. To make these contributions meaningful, I would recommend that the Dialogue be structured in a way that balances **high-level multilateral exchange with focused thematic working formats**. This could include: * **Plenary sessions** for shared strategic priorities and cross-cutting issues, * **Thematic breakout tracks** for more focused discussion, * **Written expert inputs and consultation streams** before and after the main sessions, * and **implementation-oriented roundtables** that allow stakeholders to discuss practical governance challenges rather than only broad principles. It would also be valuable to ensure that participation is not limited to those with the greatest institutional visibility or resources. The Dialogue should intentionally include voices from emerging and regulated-economy contexts, as well as practitioners working directly on implementation challenges. In short, the Dialogue will be strongest if it is designed not only to be inclusive in who participates, but also useful in how participation is structured.

Several voices remain underrepresented in global AI governance discussions, despite being directly affected by how AI is being deployed in practice. One important gap is the limited inclusion of perspectives from **emerging and regulated-economy contexts**, particularly those working in environments where institutional capacity, legal maturity, and implementation conditions differ significantly from the assumptions often embedded in dominant global governance models. A second underrepresented group includes **public-sector practitioners, institutional governance professionals, and implementation-facing experts** — those working on accountability, compliance, procurement, risk, oversight, and operational governance inside real organizations and public institutions. These actors often confront the practical governance challenges of AI directly, yet their voices are less visible than those of policymakers, major technology firms, or frontier AI researchers. There is also a need to better include perspectives from communities dealing with AI as part of **decision environments**, not only as isolated technical systems. This includes sectors and institutions where AI influences workflows, judgments, escalation paths, and public-facing decisions. These voices could be better included through: * targeted invitations and outreach, * regional and sector-specific consultation tracks, * support for written contributions from underrepresented contexts, * and formats that value implementation experience alongside formal policy or technical expertise. A globally credible AI governance dialogue must include not only those building the most advanced systems, but also those working closest to the realities of governing AI in practice.

To foster meaningful and dynamic engagement, the AI Dialogue should include formats that move beyond traditional panels and allow participants to engage with **real governance dilemmas, implementation trade-offs, and cross-sector perspectives** in a more practical way. A few formats could be especially valuable: **1) Scenario-based governance labs** Small-group sessions built around realistic AI governance cases (for example, public-sector deployment, cross-border governance conflicts, accountability failures, or human oversight breakdowns). These can help participants engage with concrete institutional challenges rather than only abstract principles. **2) Multi-stakeholder policy-design workshops** Structured sessions where governments, technical experts, academia, private sector actors, and civil society jointly work through governance design questions and produce short practical outputs or recommendations. **3) Regional and implementation roundtables** Dedicated spaces for stakeholders from different regions and institutional settings to discuss governance challenges shaped by local realities, regulatory maturity, and implementation constraints. **4) Written contribution synthesis sessions** Instead of treating written submissions as background material only, the Dialogue could include sessions that actively synthesize and respond to key themes emerging from stakeholder inputs. **5) "Challenge and response" dialogues** A format where one stakeholder group presents a real governance challenge, and others respond from their own institutional or regional perspective. This could generate more useful exchange than conventional one-way speaking formats. Overall, the most effective engagement formats will be those that prioritize **interaction, practical problem-solving, and structured exchange** over passive participation.

Several existing approaches offer useful building blocks for more effective AI governance, especially when they are treated not as isolated models, but as complementary components of a broader governance ecosystem. One important example is the **NIST AI Risk Management Framework**, which is valuable for translating high-level governance concerns into a more structured and operational risk-based approach. The **OECD AI Principles** and **UNESCO Recommendation on the Ethics of AI** also remain important because they provide globally recognized normative anchors for trustworthy, human-centered, and accountable AI governance. At the regulatory level, the **EU AI Act** is significant for showing how risk-based governance can be formalized through law, even if its implementation will continue to evolve over time. From a standards and technical governance perspective, work through **ISO/IEC** and related assurance-oriented mechanisms is also important, particularly where institutions need practical guidance for governance controls, quality assurance, and system oversight. In my view, however, some of the most effective practices are not only policy documents, but also **organizational governance mechanisms** such as: * AI risk and review committees, * human oversight protocols, * role-based accountability structures, * model and system documentation practices, * and governance checkpoints integrated into procurement, deployment, and operational workflows. The most useful lesson across these examples is that effective AI governance requires both **normative clarity and institutional implementation mechanisms**. Policies matter, but they become truly effective only when supported by governance structures that work in practice.