Alarcon & Asociados Soluciones S.A.S

In my view, the success of the first Global Dialogue on AI Governance should not be measured by the number of principles agreed upon, but by its ability to translate existing commitments into operational, accountable, and measurable frameworks. Three outcomes would signal real progress. First, convergence on implementation, not just values. There is already broad global alignment on high-level principles such as transparency, fairness, and human oversight. The challenge is operationalizing them. A successful dialogue should advance shared approaches on how these principles are embedded into system design, risk assessment, and decision-making processes—particularly in high-impact contexts. Second, clarity on responsibility across the AI lifecycle. Governance gaps often arise not from lack of regulation, but from fragmented accountability between developers, deployers, and platforms. The Dialogue should move toward clearer allocation of responsibilities, especially in relation to systemic risks and automated decision-making affecting individuals and societies. Third, strengthening mechanisms for meaningful oversight and redress. Governance is not only about preventing harm, but also about ensuring that when harm occurs, it can be identified, challenged, and remedied. This includes advancing interoperable approaches to transparency, auditability, and access to effective remedies across jurisdictions. Ultimately, the Dialogue will be successful if it helps shift AI governance from a discussion of "what should be done" to a shared understanding of how it is actually implemented, monitored, and enforced in practice.

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

My selection prioritizes the areas that are most critical to moving AI governance from principles to effective implementation. Transparency, accountability, and human oversight are essential to ensure that AI systems are not only technically robust but also explainable, contestable, and subject to meaningful human intervention-particularly in automated decision-making contexts with real impact on individuals. Without these elements, governance remains formal rather than effective. Protection and promotion of human rights provides the normative foundation for all AI governance efforts. It ensures that technological development remains aligned with internationally recognized standards, while safeguarding individuals against systemic risks and unintended harms. Safe, secure and trustworthy AI addresses the need to manage risks throughout the lifecycle of AI systems. Trust cannot be achieved through performance alone; it depends on governance structures capable of identifying, assessing, and mitigating risks in real-world deployment. Interoperability of governance approaches is critical in a global ecosystem where AI systems operate across jurisdictions. Fragmented regulatory approaches can create gaps in oversight and accountability. Promoting interoperability enables alignment between different frameworks while maintaining high standards of protection. Together, these priorities reflect a governance approach that is not limited to defining values, but focused on how AI systems are designed, deployed, supervised, and held accountable in practice.

Yes. While the listed themes cover key dimensions of AI governance, there are important cross-cutting issues that remain insufficiently captured. First, the gap between formal compliance and effective implementation. Many governance frameworks define obligations-such as transparency or risk management-but do not adequately address whether these are meaningful in practice. A system can comply procedurally while remaining opaque or unchallengeable for affected individuals. Bridging this gap requires attention to how governance measures function in real-world conditions. Second, the allocation of responsibility across the AI lifecycle. Current discussions often treat developers, deployers, and platforms separately, but in practice, accountability is distributed and sometimes fragmented. Clarifying how responsibilities are shared and enforced across actors is essential to address systemic risks and ensure effective oversight. Third, the operationalization of oversight mechanisms. Transparency, audits, and human oversight are widely recognized, yet their design and effectiveness vary significantly. Questions remain about who can access information, under what conditions, and how oversight translates into enforceable outcomes. Finally, the asymmetry between global standards and local capacity. While international principles are increasingly aligned, implementation capacity differs significantly across countries. Without addressing these disparities, governance risks reinforcing existing digital divides. These issues cut across all thematic areas and are central to ensuring that AI governance evolves from high-level commitments to effective, accountable, and context-sensitive practice.

In my context—Latin America, and particularly Colombia—governance gaps in AI are already shaping both risks and opportunities across public and private sectors. One of the most significant challenges is the gap between regulatory ambition and implementation capacity. While there is growing alignment with international standards on human rights, transparency, and accountability, institutional and technical capacities to enforce these principles remain uneven. This is particularly visible in areas such as content moderation, automated decision-making, and data governance, where oversight mechanisms are still developing. A second challenge relates to the fragmentation of responsibilities across actors. AI systems are often developed, deployed, and operated across jurisdictions, which complicates accountability when harms occur. In regions with limited regulatory coordination, this can lead to gaps in protection and enforcement. At the same time, these challenges create important opportunities. There is a strong opportunity to leapfrog towards governance-by-design approaches, integrating transparency, risk management, and human oversight early in system development rather than relying solely on ex post enforcement. This is particularly relevant for emerging digital ecosystems and public sector innovation. Additionally, the global push for interoperability of governance frameworks creates space for regions like Latin America to align with high-standard regulatory models, while adapting them to local realities. This can strengthen trust, attract responsible innovation, and enhance cross-border cooperation. Overall, the key issue is not only whether governance frameworks exist, but whether they are effectively implemented, coordinated, and enforceable in practice.

The AI Dialogue can play a critical role in advancing international cooperation by closing the persistent gap between shared principles and how they operate in practice. There is already broad global alignment on values such as transparency, accountability, and human oversight. However, the real challenge lies in ensuring that these principles translate into systems that can be understood, questioned, and effectively overseen—particularly in contexts where automated decisions directly affect individuals. In this sense, the Dialogue can function as a space to move from abstract commitments to operational clarity: how governance mechanisms are designed, how responsibilities are allocated across the AI lifecycle, and how oversight is exercised in ways that are meaningful, not merely formal. It can also strengthen interoperability across governance frameworks. AI systems operate across jurisdictions, but regulatory responses remain fragmented. Advancing compatibility between approaches is essential not only for coherence, but for ensuring that protections do not weaken when systems cross borders. Equally important is clarifying accountability. When decisions are distributed across developers, deployers, and platforms, responsibility can become diffuse. The Dialogue can help develop shared understandings of how accountability should be structured and enforced in practice. Finally, it can support more inclusive participation and capacity-building. Effective governance depends not only on rules, but on the ability of institutions and actors to implement and enforce them. Ultimately, the value of the Dialogue lies in shifting the focus from what AI governance should be, to how it is experienced in practice—especially by those affected by automated systems, who must be able not only to receive decisions, but to understand and challenge them.

The AI Dialogue should build upon existing multilateral and multi-stakeholder initiatives that have already advanced principles, standards, and early governance mechanisms, while addressing the persistent gap between coordination and implementation. Key references include the Global Digital Compact, the work of the OECD on AI principles and risk-based governance, and the UNESCO Recommendation on the Ethics of AI, which have contributed to global normative alignment. At the regional level, frameworks such as the EU AI Act and the Digital Services Act (DSA) provide more operational approaches, particularly in areas such as risk management, transparency, and accountability for high-impact systems. The Dialogue should not duplicate these efforts, but rather connect them by fostering interoperability and shared understanding across jurisdictions and governance models. One of its main added values could be to create a space where different regulatory approaches are translated into comparable practices, enabling actors to learn not only what others are doing, but how these approaches function in practice. In addition, the Dialogue can help advance coordination across the AI lifecycle, bringing together developers, deployers, regulators, and civil society to address fragmented accountability and systemic risks. Finally, it can contribute to strengthening implementation capacity, particularly in regions where governance frameworks exist but institutional and technical capabilities remain limited. Its added value, therefore, lies in moving from a landscape of multiple initiatives to a more coordinated, practice-oriented ecosystem, where governance is not only defined, but made operational, comparable, and enforceable across contexts—ensuring that governance frameworks are not only aligned, but meaningful for those affected by automated decisions.

Different stakeholders can contribute to the AI Dialogue by moving beyond representation toward functional participation aligned with their role in the AI lifecycle. Governments and regulators can provide clarity on regulatory approaches and enforcement challenges. Industry actors can contribute practical insights on system design, deployment, and risk management. Civil society and academia play a critical role in identifying impacts, testing accountability mechanisms, and ensuring that governance reflects societal needs. Technical communities can help translate governance objectives into implementable standards and tools. To be effective, the Dialogue should be structured around problem-oriented, multi-stakeholder working tracks, rather than general discussions. Each track should focus on specific governance challenges—such as risk assessment, transparency mechanisms, or accountability across the lifecycle—and aim to produce concrete, operational outputs. In terms of format, a combination of: plenary sessions to align priorities and share high-level perspectives, targeted working groups to develop practical approaches, and iterative cycles of consultation and refinement would allow both inclusiveness and depth. Importantly, participation should not be limited to formal interventions. The Dialogue should enable stakeholders to contribute evidence, case studies, and implementation experiences, creating a shared understanding of what works in practice. Finally, the structure should ensure that outputs are not only documented, but translated into actionable guidance, with clear pathways for follow-up, capacity-building, and cross-regional learning. Ultimately, the value of stakeholder participation lies in ensuring that governance is shaped not only by perspectives, but by how AI systems are designed, deployed, and experienced in practice—particularly by those affected by automated decisions.

Several voices remain underrepresented in global discussions on AI governance, particularly those most directly affected by how AI systems operate in practice. First, individuals and communities impacted by automated decision-making—in areas such as content moderation, access to services, or public sector decisions. While often labeled as "end-users," their lived experiences rarely shape governance design. Inclusion should move beyond consultation toward structured mechanisms that integrate user input into system evaluation, oversight processes, and policy development, including channels to understand, challenge, and seek remedies for automated decisions. Second, actors from regions with limited institutional and technical capacity, particularly in the Global South. Although global principles are increasingly aligned, participation in shaping governance frameworks remains uneven. Inclusion requires not only representation, but capacity-building, access to resources, and meaningful participation in decision-making processes. Third, independent researchers and civil society organizations with constrained access to data and system-level information. Effective oversight depends on access. Expanding controlled data access frameworks and supporting independent research are essential to strengthen accountability. Finally, operational and technical practitioners within organizations are often overlooked. Their insights into how systems are actually designed, deployed, and monitored are critical to making governance frameworks implementable. To address these gaps, inclusion must be designed as a structural feature of governance, not an add-on. This means embedding diverse participation into decision-making, ensuring access to relevant information, and establishing mechanisms to track how these inputs influence decisions and outcomes over time. Ultimately, more inclusive governance improves not only representation, but how AI systems are understood, governed, and experienced in practice.

Innovative engagement formats should move beyond traditional panel discussions toward interactive, problem-driven and evidence-based participation. First, the Dialogue could include scenario-based simulations, where stakeholders jointly analyze real or hypothetical cases—such as automated content moderation or AI-driven decision-making—and work through governance challenges in real time. This allows participants to move from abstract principles to practical application, revealing gaps and trade-offs. Second, multi-stakeholder co-design labs could be used to develop concrete governance tools, such as risk assessment templates, transparency models, or accountability frameworks. These formats enable collaboration across governments, industry, civil society, and technical experts, producing outputs that are directly applicable. Third, the Dialogue should incorporate evidence sessions, where participants present case studies, implementation experiences, and empirical findings. This helps ground discussions in what is actually working (or not) and supports more informed decision-making. Fourth, reverse hearings or user-centered forums could give space to individuals and communities affected by AI systems to share their experiences. This ensures that governance discussions reflect not only institutional perspectives, but also how systems are experienced in practice. Finally, iterative engagement cycles—combining in-person sessions with virtual follow-ups—can sustain dialogue over time, allowing ideas to be tested, refined, and tracked. These formats can transform the Dialogue from a space of exchange into a platform for collective problem-solving, where governance approaches are not only discussed, but co-developed, tested, and improved in practice.

Several existing policies and practices offer concrete approaches to advancing effective AI governance, particularly by moving from principles to operational mechanisms. At the regulatory level, the EU AI Act and the Digital Services Act (DSA) provide important examples of risk-based and accountability-oriented frameworks. The AI Act introduces differentiated obligations depending on the level of risk, while the DSA operationalizes governance in areas such as content moderation through requirements on transparency, statement of reasons, internal complaint mechanisms, and systemic risk assessments for very large platforms. Together, they illustrate how governance can be embedded across the lifecycle of digital systems. Beyond regulation, transparency databases and reporting mechanisms-such as platforms publishing detailed information on content moderation decisions-offer practical tools to enhance accountability. When combined with structured formats (e.g., standardized reporting or statements of reasons), these mechanisms can make decisions more understandable and contestable. Another promising practice is the development of independent audit and risk assessment frameworks, particularly for high-impact systems. These approaches help shift governance from internal compliance toward external scrutiny, even though questions remain about access, standards, and effectiveness. Additionally, multi-stakeholder governance initiatives-including OECD frameworks and UNESCO's Recommendation on AI Ethics-have contributed to aligning principles globally, while also encouraging policy experimentation at national and regional levels. However, a key lesson across these examples is that effectiveness depends not only on the existence of rules, but on how they are implemented, monitored, and enforced in practice. Approaches that integrate transparency, accountability, and oversight into system design-rather than treating them as afterthoughts-are more likely to produce meaningful governance outcomes. Ultimately, good practices are those that make governance not only visible, but actionable and responsive to real-world impacts.