i-visionet

A successful first Global Dialogue on AI Governance must confront a fundamental imbalance: AI is globally deployed, but governance remains concentrated in a small number of countries that control infrastructure, data, and standards. Without addressing this, any global framework will remain incomplete. First, the Dialogue should formally recognize that countries without control over AI infrastructure—such as the Democratic Republic of Congo—face structural limitations in governing AI systems. This is not a marginal issue; it reflects the reality of a large part of the world. Second, it should advance practical governance mechanisms at the point of deployment, including pre-deployment testing, risk-based validation, and transparency requirements. These are essential tools for countries that cannot exercise upstream control over models and data. Third, the Dialogue must acknowledge that AI governance is also linked to energy, resources, and data value chains. Countries like the DRC, which are central to these ecosystems, must be included not only as users, but as stakeholders in shaping governance frameworks. Finally, success requires recognizing that supporting governance capacity in such contexts is a global priority. Weakly governed environments create systemic risks, while well-supported ones can serve as real-world testing grounds for more robust and globally applicable governance models. Success lies in building governance that reflects these realities—not ignoring them.

• AI capacity-building
• Transparency, accountability, and human oversight
• Interoperability of governance approaches
• Social, economic, ethical, cultural, linguistic and technical implications of AI

These priorities reflect the need to ground global AI governance in the realities of countries that are structurally constrained in terms of infrastructure, data ecosystems, and institutional capacity. AI capacity-building is essential because effective governance cannot exist without the minimum technical, regulatory, and institutional capabilities to understand and oversee AI systems. For countries like the Democratic Republic of Congo, this is a prerequisite for meaningful participation in global governance, not a secondary objective. The social, economic, ethical, cultural, linguistic and technical implications of AI are central, as AI systems deployed globally often fail to reflect the contexts in which they operate. Linguistic diversity, informal economies, and socio-cultural dynamics directly affect system performance, fairness, and safety. Ignoring these factors introduces systemic risks and limits the reliability of AI at scale. Interoperability of governance approaches is critical to ensure that global frameworks are not limited to a narrow set of high-capacity environments. Governance models must be adaptable and capable of functioning across diverse regulatory and operational contexts, including those where upstream control over infrastructure is limited. Finally, transparency, accountability, and human oversight are indispensable in environments where AI systems are developed externally. Without visibility into how systems operate, countries cannot effectively manage risks, detect bias, or ensure alignment with national priorities. Together, these priorities support the development of practical, inclusive, and globally applicable AI governance frameworks that reflect the realities of all countries, not only those with advanced technological capabilities.

While the identified themes provide a comprehensive foundation for global AI governance discussions, an important cross-cutting dimension could be further emphasized: the interdependence between AI systems, underlying infrastructure, and global resource value chains. The ability of States to effectively govern AI is closely linked not only to regulatory frameworks, but also to access to infrastructure, compute capacity, energy, and the resources that enable the development and deployment of AI technologies. These structural factors influence how AI systems are developed, deployed, and governed across different contexts. In this regard, there is growing attention to the risk of uneven value distribution within the global AI ecosystem, particularly for countries that contribute significantly through natural resources or data, but may have limited participation in downstream value creation. This dynamic is sometimes described as forms of "digital extraction," where the benefits of AI development are not equitably shared. Further reflection on how global governance frameworks can better account for these interdependencies-while promoting transparency, fairness, and inclusivity-could enhance their relevance and effectiveness. Integrating considerations related to infrastructure, resource governance, and value distribution would support a more holistic and globally representative approach to AI governance.

In the context of the Democratic Republic of Congo, governance gaps in the selected thematic areas are closely linked to structural constraints, particularly in relation to infrastructure, data ecosystems, and institutional capacity. One of the most significant challenges is the limited ability to exercise effective oversight over AI systems developed and hosted externally. This affects transparency, accountability, and the capacity to assess risks such as bias, reliability, and alignment with national priorities. In addition, linguistic and socio-economic diversity presents challenges for the safe and effective deployment of AI systems, which are often not designed for such contexts. Another key challenge relates to capacity-building, including the availability of technical expertise, regulatory frameworks, and coordination mechanisms required to govern AI effectively. Without these, governance risks remaining largely theoretical. At the same time, these gaps also present important opportunities. They create space to develop context-adapted governance approaches, including deployment-level oversight mechanisms and risk-based frameworks that are more suited to infrastructure-dependent environments. Furthermore, the country's role in energy and resource value chains relevant to AI provides an opportunity to contribute to broader discussions on sustainability, value distribution, and the future of global AI ecosystems. More broadly, addressing these governance gaps could enable the development of practical and inclusive governance models that are applicable not only nationally, but across similar contexts globally, thereby strengthening the overall resilience and relevance of international AI governance efforts.

The AI Dialogue can play a pivotal role as a platform for building shared understanding and trust across diverse national contexts, which is essential for effective international cooperation on AI governance. First, it can facilitate recognition of structural differences among countries, including disparities in infrastructure, data ecosystems, and institutional capacity. Acknowledging these realities is critical to ensuring that global governance approaches are inclusive and applicable across different environments. Second, the Dialogue can support the exchange of practical experiences and approaches, enabling countries to share context-specific solutions, including governance mechanisms adapted to infrastructure-dependent or resource-constrained settings. This can contribute to the development of more flexible and interoperable frameworks. Third, it can promote co-development of governance principles and tools, rather than one-directional adoption. By bringing together a wide range of stakeholders, the Dialogue can help ensure that emerging standards reflect diverse perspectives, including those of countries that are not primary developers of AI technologies but are significantly impacted by them. Finally, the AI Dialogue can help mobilize international cooperation and support, particularly in areas such as capacity-building, technical expertise, and institutional strengthening. Strengthening governance capabilities in all regions contributes to reducing systemic risks and enhancing the overall stability and reliability of the global AI ecosystem. In this way, the Dialogue can serve as a foundation for more inclusive, coordinated, and context-aware international AI governance.

The AI Dialogue can play a pivotal role as a platform for building shared understanding and trust across diverse national contexts, which is essential for effective international cooperation on AI governance. First, it can facilitate recognition of structural differences among countries, including disparities in infrastructure, data ecosystems, and institutional capacity. Acknowledging these realities is critical to ensuring that global governance approaches are inclusive and applicable across different environments. Second, the Dialogue can support the exchange of practical experiences and approaches, enabling countries to share context-specific solutions, including governance mechanisms adapted to infrastructure-dependent or resource-constrained settings. This can contribute to the development of more flexible and interoperable frameworks. Third, it can promote co-development of governance principles and tools, rather than one-directional adoption. By bringing together a wide range of stakeholders, the Dialogue can help ensure that emerging standards reflect diverse perspectives, including those of countries that are not primary developers of AI technologies but are significantly impacted by them. Finally, the AI Dialogue can help mobilize international cooperation and support, particularly in areas such as capacity-building, technical expertise, and institutional strengthening. Strengthening governance capabilities in all regions contributes to reducing systemic risks and enhancing the overall stability and reliability of the global AI ecosystem. In this way, the Dialogue can serve as a foundation for more inclusive, coordinated, and context-aware international AI governance.

Effective contribution to the AI Dialogue requires inclusive, structured, and multi-level engagement across stakeholders. Different stakeholders can contribute in complementary ways. Member States can articulate national priorities, regulatory experiences, and context-specific challenges. Private sector actors can provide technical insights, operational practices, and implementation perspectives. Academia and research institutions can contribute evidence-based analysis and evaluation methodologies. Civil society can ensure that societal, ethical, and human rights considerations are fully reflected. International organizations can facilitate coordination, knowledge-sharing, and alignment with existing frameworks. To maximize impact, the Dialogue could adopt a layered and iterative structure: Thematic working groups aligned with priority areas (e.g., capacity-building, transparency, interoperability), allowing for focused and sustained engagement. Regional or context-based sessions, enabling countries with similar structural conditions to share practical approaches and identify common challenges. Technical and policy exchanges, bridging high-level principles with operational realities. Periodic synthesis outputs, capturing areas of convergence, key gaps, and actionable recommendations. In terms of format, combining formal plenary discussions with smaller, interactive formats (roundtables, expert panels, and case-based exchanges) would encourage both strategic dialogue and practical knowledge-sharing. Importantly, the Dialogue should remain continuous rather than one-off, allowing for iterative learning, refinement of approaches, and progressive alignment over time. Such a structure would support co-creation, mutual understanding, and the development of governance approaches that are both globally coherent and adaptable to diverse national contexts.

Global discussions on AI governance have made important progress toward inclusivity; however, certain perspectives remain underrepresented. In particular, countries with limited AI infrastructure and constrained institutional capacity, including many in the Global South, often have fewer opportunities to shape governance frameworks, despite being significantly impacted by their outcomes. In addition, communities operating in linguistically diverse and low-connectivity environments are not always adequately reflected in current discussions. Their experiences are critical to understanding how AI systems function in real-world contexts that differ from those in which they are primarily developed. Perspectives from sectors closely linked to global resource and data value chains, including those contributing to energy and critical minerals essential to AI, also warrant greater attention. These stakeholders can provide important insights into issues related to sustainability, value distribution, and long-term resilience of the AI ecosystem. To enhance inclusion, several approaches could be considered. Expanding regional consultations and context-specific dialogues would allow for more representative participation. Supporting capacity-building and technical assistance can enable a broader range of stakeholders to engage meaningfully. Additionally, ensuring that dialogue formats accommodate diverse forms of participation, including multilingual engagement and hybrid or remote access, can help reduce barriers. Strengthening the participation of these perspectives would contribute to more balanced, applicable, and globally representative AI governance frameworks.

To foster meaningful and dynamic engagement, the AI Dialogue could combine traditional plenaries with interactive, problem-oriented formats that bridge policy and practice. First, case-based workshops would allow participants to examine concrete scenarios—such as AI deployment in low-connectivity or linguistically diverse environments—and collectively identify governance responses. This grounds discussions in real-world conditions and encourages practical solutions. Second, multi-stakeholder "co-creation labs" could bring together governments, private sector actors, researchers, and civil society to jointly design policy approaches or tools (e.g., risk assessment templates or deployment guidelines). These sessions would emphasize collaboration rather than one-directional exchanges. Third, regional and context-specific dialogues would enable participants from similar environments to share experiences and identify common challenges, particularly for countries with infrastructure constraints or emerging governance frameworks. Fourth, technical-policy bridging sessions could connect high-level governance discussions with operational realities, allowing technical experts to explain system limitations while policymakers articulate regulatory expectations. In addition, interactive formats such as moderated roundtables, scenario simulations, and peer-learning exchanges can promote more open and balanced participation, especially for stakeholders who may be less represented in formal plenary settings. Finally, the Dialogue could include periodic synthesis sessions to capture key insights, areas of convergence, and actionable recommendations, ensuring continuity and impact beyond individual discussions. Such a combination of formats would support inclusive, practical, and context-aware engagement, enabling participants to move from principles to actionable governance approaches.

A range of existing policies and practices provide useful foundations for effective AI governance, particularly when adapted to different national contexts. At the international level, the principles developed by the OECD have helped establish a common baseline around transparency, accountability, and human-centered values. Similarly, the risk-based approach advanced by the European Commission through its AI regulatory framework offers a structured method for classifying and managing AI risks according to their potential impact. From a practical perspective, the AI Risk Management Framework developed by the National Institute of Standards and Technology provides actionable guidance for identifying, assessing, and mitigating risks across the AI lifecycle. This type of framework is particularly useful for translating high-level principles into operational processes. In addition, regulatory sandboxes have emerged as a pragmatic tool, allowing governments to test AI systems in controlled environments before wider deployment. While not a complete solution, they offer a structured entry point for oversight, especially in contexts where upstream visibility into systems is limited. Other promising approaches include the development of national data governance strategies, the use of public procurement standards to enforce accountability and transparency requirements, and the promotion of multistakeholder platforms that enable collaboration across governments, industry, and civil society. Taken together, these examples demonstrate that effective AI governance often relies on combining principles, risk-based frameworks, and practical implementation tools, adapted to local capacities while aligned with global standards.