South Centre

The measure of success for the first Dialogue will be in the advancement of a global agenda for AI governance that improves regulatory coherence, increases access and benefit-sharing of AI, and generates positive AI impacts for sustainable development in the global South. First, the global AI governance regime is currently fragmented, with different processes leading to a plurality of outcomes that frequently lack coherence, show redundancy, and are not being effectively implemented. Drawing on the work of the UN AI Scientific Panel, the Dialogue should enable stakeholders to identify commonalities and synergies among the different AI governance initiatives at national, regional and multilateral levels. Further, it should advance complementarity across institutional frameworks, such as within the UN system, African Union and BRICS, among others. Second, the Dialogue should consider how AI access and capacity-building can be enhanced to address the structural barriers faced by global South countries in AI development, deployment and governance. This requires action for promoting access to digital infrastructure, compute resources, representative datasets, and AI skills development. It should identify resources that can be engaged, such as the Global Fund for AI and complementary initiatives such as implementation of UN Resolution 78/311. Third, the Dialogue should consider the distributional impacts of AI on developing and least developed countries. The majority of global South countries and stakeholders are being impacted by AI but remain largely unable to influence how this technology is advancing and being applied. The New Delhi Declaration on AI Impact rightly emphasised the need to democratise AI resources and foster inclusive multilateral engagement. Advancing the voice of the global South in the Dialogue requires their effective participation in its agenda setting and discussions, bringing their concerns and priorities to the fore.

• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Open-source software, open data and open AI models

AI capacity-building is critical as developing nations tend to lack the infrastructure, skills, and governance frameworks needed to benefit from AI-related productivity gains. Addressing this requires technology transfer, financing for digital infrastructure, and AI skill development. This should be supported through increasing investment in developing AI literacy, supporting research and innovation, and integration into productive industrial sectors with already high digitalisation. The Dialogue should help realise this by coordinating action among all stakeholders at different levels. The multidimensional implications of AI must be considered as increasing AI adoption will have differentiated impacts on global South countries. The lack of representation in AI models is clearly reflected in current AI training datasets which are heavily skewed toward advanced economies. This produces biased and suboptimal outcomes when these AI models are deployed in the global South. The Dialogue can advance sustainable AI solutions from developing countries which are pertinent to their specific contexts, such as specific linguistic needs and limited resource availability, and can be shared through South-South cooperation modalities. Realizing the full and effective enjoyment of all human rights must be embedded into AI governance. AI can expand existing inequalities, which would adversely impact the most vulnerable groups, especially in the global South. For promoting the GDC's objective on protecting human rights in the digital space, the Dialogue should focus on integrating transparency, accountability and robust human oversight of AI for reducing biases and data extractivism, and supporting access to safe, secure and trustworthy AI. Finally, prioritising open-source in AI can help developing countries in addressing concerns around AI access and technological dependency. Open-source AI can enable countries to use AI models and adapt them to their national needs, contexts and priorities. It reduces dependence on foreign Big Tech companies and aligns with the GDC's focus on developing open AI resources.

1. The issue of data sovereignty and equitable benefit-sharing from data should be considered by the Dialogue. Developing countries' data have been utilised in training AI models, yet the countries and people generating such data have received little or no benefit from such extraction and use. Complementing the work of the UN Working Group on Data Governance, the Dialogue should discuss how equitable data governance, access and benefit-sharing can be promoted in the specific context of AI systems, in a manner that safeguards developing countries' sovereign rights over their data. 2. The sustainable development impacts of AI should be articulated and discussed by the Dialogue, informed by the report of the AI Scientific Panel. For e.g. with 1.2 billion people expected to reach working age in developing countries in the next decade, providing decent and high-quality jobs is a critical challenge with deep impacts on sustainable development pathways, and directly related to attainment of several SDGs. The Dialogue should focus on supporting global South countries to reduce adverse AI impacts, and ensure that AI augments rather than impedes sustainable development. 3. While multistakeholder perspectives should inform the Dialogue, it is important that AI regulation and governance maintain the government as the primary actor. The emphasis of AI governance is broad, encompassing access, innovation, infrastructure, safety and application, but should keep overall public interest as a paramount objective. Learning from the experiences of some global South countries in developing and deploying digital public infrastructure (DPI) as a public good is useful in this context. Signalling support from the Dialogue for building publicly-owned and jointly developed open AI systems and infrastructure would be useful, as this would safeguard data sovereignty, reduce dependence on foreign and proprietary AI models, and foster greater collective AI capacities among developing countries.

For most global South countries, there are many gaps in relation to AI governance, with potentially significant adverse impacts across different domains and on their sustainable development. The extensive fragmentation in global AI governance makes it difficult for developing countries with limited resources to engage meaningfully across the many fora where AI governance is being discussed. This also produces overlapping outcomes and competing agendas, further weakening effective implementation. With significant asymmetry between developed and developing countries in their participation in international norm-setting processes, the priorities and perspectives of the global South are not being sufficiently reflected in multilateral deliberations. At the national level, gaps in data governance, whether in legislation or enforcement, have led to raw data being routinely extracted from developing countries, local communities and global South stakeholders without their informed consent or compensation. Without comprehensive data governance frameworks and mechanisms to enforce them, especially in the context of cross-border data flows, developing countries risk being locked into a dependent role in the AI economy as sources for data extraction. In the area of AI infrastructure, developing countries are failing to effectively compete with advanced economies for securing access to the compute resources such as new generation GPUs which are necessary for AI development and deployment. Unable to mobilise sufficient resources at the scale required, they are having to rely on voluntary initiatives by external partners for accessing AI technologies, financing and building national capacity. This risks creating new technological dependencies for developing nations without meaningfully bridging their existing digital divides. The Dialogue thus offers an important opportunity to proactively address these gaps and enable global South nations to meaningfully contribute to global AI governance before AI entrenches existing hierarchies and widens inequalities within and across countries.

The Dialogue has the mandate to convene the full range of stakeholders for advancing international cooperation on AI governance. To achieve this, it should first serve as a platform for facilitating coordination among all the stakeholders pursuing different AI governance processes globally. This includes bringing together the UN system, regional bodies such as the African Union and ASEAN, international organisations, think-tanks, academia, private sector and civil society from across countries and regions. Further, the Dialogue should leverage the legitimacy conferred by its status as a UN mandated initiative to increase coherence among existing initiatives on AI governance, including perspectives and standards emerging from the global South. For instance, the 2025 BRICS Statement on the Global Governance of AI and the G20 AI for Africa Initiative both reflect the need for an inclusive, UN-based process. The Dialogue should build on the momentum generated by the India-AI Impact Summit to democratise AI technologies, identify common concerns around AI development and deployment, as well as ways for developing common baselines for governing AI, especially in relation to its safety and trust. The Dialogue should further consider advancing international cooperation in specific areas related to AI, such as enhancing access, skills and capacity-building. For instance, this could include facilitating the realisation of objectives under the GDC and supporting initiatives under UN Resolution 78/311. Further, it could foster international partnerships for access to AI technologies and compute, and mobilise resources for the Global Fund for AI, which would promote responsible AI adoption in global South nations.

The Dialogue should remain firmly grounded in the GDC and its implementation framework, serving as an inclusive space within the UN framework to discuss the most pressing AI challenges. To take the mandate of UN Resolution 79/325 forward in the most inclusive and participatory manner, the Dialogue should build on existing outcomes and process that have advanced global South priorities, such as the AI for Africa Initiative and the India-AI Impact Summit. For instance, it could take inspiration from the New Delhi Declaration on AI Impact and the Charter for the Democratic Diffusion of AI to identify how its priorities can be effectively translated and operationalised at the multilateral level, especially by the UN system. The added value of the Dialogue could be realised through positioning it as the multilateral platform for advancing AI governance in alignment with global sustainable development priorities. The AI Scientific Panel will provide the Dialogue with policy-relevant views on the development of AI and its impacts on countries and societies. This should first be widely disseminated by the Dialogue and second, used as a starting point to facilitate open, transparent exchanges across governments, policymakers, and all other relevant stakeholders for supporting meaningful and effective global AI governance. In this manner, the Dialogue could help to address the concerns around AI's multidimensional impacts and governance gaps that disproportionately affect developing countries.

The Dialogue must enable effective and meaningful participation by diverse stakeholders, especially from across the global South. Given the primacy of governments in the regulation and governance of AI, countries should be given dedicated space for sharing their knowledge, experiences and lessons learned. This would enable better understanding of national priorities and governance approaches for AI, as well as to better identify good practices that could be replicated in other contexts, especially for AI safety and trust. The inputs of non-governmental stakeholders are critical as they are on the frontlines and can give first-hand evidence of how AI is being deployed and used on the ground, and what impacts it is having on economies and societies. Stakeholders such as civil society organisations, academics, private sector representatives, and technical communities should be able to contribute to the Dialogue through submitting inputs which document their experiences, perspectives and technical know-how. Taking forward the India-AI Impact Summit's use of casebooks and showcasing use cases for AI solutions, the Dialogue could spotlight successful AI governance initiatives and practices from developing countries which can be replicated and adopted by others. Using the evidence base so generated, international organisations, research institutions and think-tanks could analyse trends, provide insights and suggest actionable policy recommendations which might be useful from the perspective of the global South. Finally, enabling access to the Dialogue through a virtual platform and livestreaming of its sessions would further expand its global reach and participation. The first Dialogue should set a clear precedent for responsible and inclusive multistakeholder engagement in global AI governance and for integrating diverse perspectives, especially from the global South, into its outcomes.

The current global discourse on AI governance has focused on the stated priorities of a small handful of big tech companies and large economies which have the infrastructure capacities, technical know-how and financial resources to develop cutting-edge AI systems. Global South voices have been rather marginalised in AI governance discussions till quite recently. Stakeholders based in global South countries have little agency in shaping how AI is being developed and deployed globally. Certain communities have even higher vulnerabilities, for instance, workers such as gig workers and data labellers interact regularly with AI systems, yet have no influence on how AI adoption is changing their working conditions. Even in the private sector, small businesses and MSMEs are relegated to being passive recipients of AI. To include these stakeholders' perspectives, the Dialogue should give higher weight to the evidence and inputs from developing countries researchers, institutions and communities. It should proactively consider facilitating the travel of global South stakeholders to Geneva and New York for providing their insights. This should be complemented through convening national and regional consultations in Asia, Africa and Latin America and Caribbean that could better inform the Dialogue and make it inclusive and equitable for all. Further, there are noted biases in global AI discussions towards considering only the positive potential of AI and minimising its downsides. The Dialogue should correct this by engaging with stakeholders that are critical of AI technologies and highlight their adverse impacts. This would result in a more balanced and rigorous process for strengthening the Dialogue and global AI governance as a whole.

For advancing global AI governance, the Dialogue should invite countries and policymakers to showcase their national AI policies and innovations, for instance, through the use of case studies covering the successful deployment of AI solutions in different policy areas. This would highlight the specific impacts of AI-based policy interventions, the lessons learned, and implementation challenges faced. The Dialogue should also enable other stakeholders to present the challenges and opportunities arising from their own development and deployment of AI, its grassroots impacts, as well as the approaches taken to dealing with them. Taking inspiration from the Global Digital Public Infrastructure Repository established during India's G20 Presidency, the Dialogue could create a repository of replicable and scalable AI solutions that promote sustainable development. Such a repository would enable better cataloguing and discovery of AI solutions, particularly those well-suited for developing country contexts. The Dialogue should also consider a wider time horizon going beyond the anticipated two-day event in July 2026 and in 2027. For instance, it could catalyse inter-sessional work by establishing subsidiary multistakeholder working groups on the specific thematic priorities and areas identified in Resolution 79/325. These working groups could have inter-sessional meetings, enable more focused engagement, and produce actionable insights that would then feed into the Dialogue itself. It would be necessary to maintain appropriate geographical and sectoral representation in these groups, considering a wide membership comprising representatives from governments, civil society, academia, and technical communities, with active involvement of stakeholders from the global South. Finally, enabling virtual means of engagement for stakeholders that are unable to attend in person would dramatically expand inclusion of diverse voices. This would help ensure the Dialogue genuinely reflects the diversity of the global AI community.

Several global South countries and regional bodies have pioneered innovative AI governance initiatives, elaborated below: The African Union focuses on continent-level coordination with policy instruments like the Continental Strategy on Artificial Intelligence and the African Continental Free Trade Agreement (AfCFTA) Protocol on Digital Trade situating AI governance within the Agenda 2063, emphasising harmonisation of data governance, interoperability, and shared infrastructure for a unified digital ecosystem. The ASEAN Guide on AI Governance and Ethics provides national and regional level recommendations and aims to empower organisations and governments in the region to design, develop, and deploy AI systems responsibly. It includes a voluntary AI Governance Framework with principles for responsible AI, emphasizing innovation-friendliness and relevant social context of South-East Asian countries. Brazil's AI regulatory sandbox regime, administered by the National Data Protection Authority (ANPD), focuses on creating a controlled environment for testing AI technologies, generating evidence on risks and impacts, and informing regulation while fostering innovation. Furthermore, the Brasil Mais Produtivo programme demonstrates how MSMEs can be supported to increase productivity through AI and digital transformation. China's AI governance framework is composed of several technical and administrative directives that address different aspects of AI. Its 'AI Plus' initiative aims to rapidly diffuse AI into key areas, including scientific research, technological innovation, and improving governance. It is an ambitious plan that includes a target of integrating AI into 90 percent of China's economy by 2030. India's AI Governance Guidelines seek to advance technical progress and mitigate the potential risks of AI to society, while being grounded in national needs and aspirations. India recently proposed a techno-legal AI governance framework which would embed safeguard mechanisms directly at the design and development stage of AI models, thereby enabling compliance for safe and trustworthy AI and promote responsible AI by design.