A2V - AI Center of Excellence

A successful first Global Dialogue on AI Governance should deliver three outcomes. First, conceptual alignment with operational clarity. There is broad agreement on principles for safe, secure and trustworthy AI. The gap lies in translating these into implementable governance structures. The Dialogue should therefore move beyond high-level alignment and define how governance, accountability and security interact in practice. Second, integration of AI governance and cybersecurity. In many institutions and organizations, these remain separate domains. In practice, however, they are interdependent. Without integrating security considerations into governance frameworks from the outset, systems risk being deployed that are difficult to secure, oversee and trust. Addressing this structural disconnect should be a priority outcome. Third, a focus on institutional capacity. The next divide in AI will not be technological, but institutional. Many countries and organizations are engaging with AI, but lack the capacity to implement governance effectively. The Dialogue should therefore lead to concrete capacity-building approaches, particularly for emerging and developing regions. To be meaningful, the Dialogue should also produce practical outputs: a shared understanding of priority implementation gaps, initial guidance for institutions, and a roadmap for continued cross-regional cooperation. This would ensure that the Dialogue contributes not only to discussion, but to operational readiness and trust in AI systems globally.

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

The selected priorities reflect the need to move from principles to implementation. Safe, secure and trustworthy AI is the overarching objective. However, trust cannot be achieved through design principles alone. It depends on whether governance and security are effectively embedded in real-world systems. AI capacity-building is critical, as the main gap is not awareness, but the ability of institutions to implement governance in practice. This is particularly relevant in emerging and developing regions, where adoption is accelerating but institutional frameworks are still evolving. Interoperability of governance approaches is essential in a global context. Fragmented frameworks risk creating inconsistencies, regulatory friction and gaps in oversight. Greater alignment does not require uniformity, but coordination and mutual understanding across regions. Transparency, accountability and human oversight form the foundation of responsible AI deployment. In practice, however, these are often not clearly defined at the level of operational responsibility. Clarifying ownership of risk and decision-making is therefore a key requirement. Across all four priorities, one issue is consistently visible in practice: AI governance and cybersecurity are still treated separately. Addressing this gap would significantly strengthen all selected areas and contribute directly to building systems that institutions and societies can trust.

A key cross-cutting issue is the growing gap between AI adoption and institutional governance capacity. Across regions, organizations are rapidly deploying AI systems, while governance structures, risk ownership and security integration are often not evolving at the same pace. This creates a structural imbalance: systems scale faster than they can be effectively governed and secured. A second emerging issue is the disconnect between policy and implementation. While regulatory and ethical discussions are advancing, institutions frequently lack the operational models, skills and structures required to apply these frameworks in practice. A third issue is the separation of AI governance and cybersecurity. In many environments, these are treated as distinct domains, which leads to systems being developed without integrated consideration of resilience, threat exposure and long-term oversight. This limits the ability to build trustworthy systems. Finally, there is a need to address cross-regional asymmetries. Different regions are progressing at different speeds in terms of regulatory maturity, technical capacity and institutional readiness. Without structured mechanisms for knowledge transfer and capacity-building, this may lead to fragmented governance landscapes. Addressing these issues requires a stronger focus on implementation, institutional capability and cross-domain integration, rather than additional high-level principles.

Across Latin America, the most significant dynamic is a growing gap between AI adoption and institutional governance capacity. On the opportunity side, there is strong momentum. Organizations across the private sector, public institutions and academia are actively experimenting with AI. There is openness to innovation, increasing awareness of governance topics, and a willingness to engage in international dialogue. However, three structural challenges are consistently visible. First, limited institutional capacity. While interest in AI governance is high, many organizations lack the structures, skills and processes required to define risk ownership, ensure accountability and implement oversight in practice. Second, fragmentation between AI governance and cybersecurity. These are often treated as separate domains, which leads to systems being deployed without integrated consideration of resilience, threat exposure and long-term security. As a result, risks are addressed reactively rather than by design. Third, misalignment between policy discussions and operational reality. Regulatory and ethical frameworks are advancing, but institutions frequently struggle to translate them into implementable models. This creates uncertainty and slows down effective governance. At the same time, this situation creates a clear opportunity. Regions such as Latin America can build governance models that integrate AI governance, cybersecurity and institutional capacity from the outset, rather than adapting fragmented legacy approaches. This would allow for more coherent, scalable and trustworthy systems, and position the region as an active contributor to global governance discussions. Addressing these challenges requires a stronger focus on implementation, cross-domain integration and capacity-building, supported by international cooperation.

The AI Dialogue can play a critical role as a bridge between global principles and institutional implementation. At present, there is no shortage of frameworks, guidelines or initiatives on AI governance. The main challenge is fragmentation and the lack of translation into operational reality across different regions. In this context, the Dialogue can create value in three ways. First, as a convergence platform. It can bring together governments, private sector, academia and technical communities to align on key governance concepts, while respecting regional diversity. This includes fostering interoperability between governance approaches rather than uniformity. Second, as a connector between domains. A recurring gap in practice is the separation between AI governance and cybersecurity. The Dialogue can explicitly address this by promoting integrated approaches that link governance, risk management and security considerations. Third, as an enabler of implementation. Beyond discussion, the Dialogue should support the translation of principles into practice by identifying priority gaps, sharing applied models and enabling cross-regional capacity-building. In this sense, the Dialogue should not be seen only as a forum for exchange, but as a mechanism to accelerate institutional readiness. Its success will depend on whether it helps stakeholders move from alignment on principles to coordinated action and implementation across regions.

The AI governance landscape is already shaped by a range of important initiatives, including national and regional regulatory frameworks, multilateral processes, and standards-setting bodies. These efforts provide a strong foundation, particularly in areas such as risk-based regulation, human rights considerations and technical standards. However, they often evolve in parallel and are not always connected in a way that supports consistent implementation across regions. In addition, many initiatives focus either on policy design or on technical development, while the institutional and operational layer remains underdeveloped. The added value of the AI Dialogue lies in its ability to connect these existing efforts and address the implementation gap. Specifically, the Dialogue could: - create structured links between policy frameworks and operational practices, - facilitate exchange on applied governance models across regions, - promote interoperability between different governance approaches, - and support capacity-building efforts aligned with real institutional needs. A further opportunity is to explicitly integrate AI governance and cybersecurity perspectives, which are often addressed separately in existing initiatives. By acting as a coordination and translation platform, the Dialogue can help ensure that existing work leads to coherent, scalable and practically applicable governance models. This would strengthen international cooperation not only at the level of principles, but at the level of implementation and institutional capability.

Effective participation in the AI Dialogue requires moving beyond representation towards structured contribution and co-creation. Different stakeholders bring complementary strengths: - Governments provide regulatory direction and institutional frameworks. - Private sector actors contribute implementation experience and operational insights. - Academia and technical communities ensure methodological rigor and technological depth. - Civil society brings critical perspectives on societal impact and trust. To make this diversity meaningful, the Dialogue should be structured around joint workstreams rather than parallel discussions. In particular, cross-domain collaboration is essential. A recurring gap in practice is the separation between AI governance, cybersecurity and operational implementation. Structuring the Dialogue around integrated themes would enable more coherent outcomes. From a format perspective, a combination of: - focused thematic working groups, - cross-regional roundtables, - and synthesis sessions translating discussion into concrete outputs would be effective. In addition, continuity is critical. The Dialogue should not be a one-off exchange, but part of an ongoing process of knowledge-sharing, capacity-building and implementation support across regions. In practice, stakeholders who are already working across sectors and regions can contribute by helping to translate global principles into operational models, ensuring that discussions remain connected to real-world implementation.

A key underrepresented perspective in global AI governance discussions is that of institutions and practitioners working at the implementation level, particularly in emerging and developing regions. While global discussions are often shaped by policy-makers, large technology actors and research institutions, there is comparatively less visibility of those responsible for applying governance frameworks in practice within organizations, public institutions and local ecosystems. This includes: - public sector entities managing AI adoption under resource constraints, - small and medium-sized enterprises implementing AI without dedicated governance structures, - and cross-functional practitioners bridging policy, technology and security domains. In regions such as Latin America, where I am operating mostly, there is strong engagement with AI, but limited representation in shaping global governance narratives. This creates a risk that frameworks are developed without sufficient consideration of local institutional realities. To address this, inclusion should go beyond participation and focus on structured integration of practical experience. This could include: - targeted inclusion of practitioners from diverse regions, - dedicated sessions focused on implementation challenges, - and mechanisms to systematically capture and reflect operational insights in the Dialogue outputs. Strengthening these perspectives would improve the practical relevance, inclusiveness and global applicability of AI governance frameworks.

To foster meaningful engagement, the Dialogue should combine inclusiveness with structured, outcome-oriented formats. Traditional panel discussions often allow for broad exchange, but limited depth and continuity. Complementary formats could enhance effectiveness. First, implementation-focused working sessions. Small, diverse groups could work on specific governance challenges, such as integrating AI governance and cybersecurity, or defining operational accountability models. These sessions should aim to produce concrete outputs rather than general discussion. Second, cross-regional exchange formats. Structured dialogues between regions at different stages of AI adoption could facilitate knowledge transfer and highlight diverse institutional realities. Third, case-based discussions. Real-world examples of AI deployment and governance challenges can provide a practical anchor, enabling participants to move beyond abstract principles. Fourth, iterative engagement. The Dialogue should include follow-up mechanisms, allowing stakeholders to refine ideas, share progress and build on previous discussions over time. Across all formats, the key objective should be to connect policy, technology and implementation perspectives. This would ensure that the Dialogue not only captures diverse views, but also contributes to developing actionable, globally relevant governance approaches.

Effective AI governance is increasingly shaped by approaches that combine risk-based frameworks, operational accountability and cross-domain integration. A first important development is the adoption of risk-based governance models, which allow institutions to differentiate between use cases and apply proportional oversight. These approaches provide a strong foundation, but their effectiveness depends on how they are implemented within organizations. A second emerging practice is the operationalization of accountability. Defining clear ownership of risk, decision-making and oversight at the institutional level is critical. In practice, this often requires embedding governance into existing processes rather than treating it as a separate compliance layer. A third area of growing importance is the integration of AI governance and cybersecurity. In many environments, these remain separate functions. However, effective governance increasingly requires aligning policy frameworks with security architecture, ensuring that resilience, threat exposure and system integrity are considered from the outset. In addition, capacity-building approaches that combine technical, governance and organizational perspectives are proving essential. Institutions require not only guidelines, but also the skills and structures to apply them. From a practical perspective, effective approaches often share three characteristics: - they are embedded in institutional processes, not external to them, - they connect governance, security and operations, - and they are designed to be adaptable across different regional contexts. These elements are critical to moving from principles to operational, trustworthy AI systems. I would welcome the opportunity to contribute to this process and support the translation of global principles into practical implementation across regions.