The Global Partnership for Sustainable Development Data

The first Global Dialogue will be successful if it lays the groundwork for practical consensus among stakeholders on the core building blocks of inclusive, effective AI governance. Success should be measured based on the extent to which the first Dialogue clarifies shared priorities, identifies actionable areas for cooperation, and connects high-level governance discussions to the practical realities faced by stakeholders. In our view, a successful outcome would include convergence around several key points, including: (1) that trustworthy AI depends on access to high-quality, well-governed data (2) and that AI governance must address the full lifecycle of AI systems, including the data practices that underpin them. The Dialogue should set a foundation for ongoing collaboration among governments, multilaterals, civil society, researchers, the private sector, and the general public. It should elevate implementation challenges faced by countries and institutions with fewer resources and ensure that governance discussions are informed by the daily realities of public-interest data use, digital public infrastructure, and institutional capacity. Success would mean moving the global conversation on AI from abstract principles toward a grounded, cooperative agenda for AI governance that is equitable, development-oriented, and responsive to the range of stakeholder concerns across diverse contexts. In addition to substantive outcomes, a successful AI Dialogue also requires a process that is inclusive, transparent, and meaningfully participatory. The legitimacy of the Dialogue's outcomes will depend on who has power to shape those decisions. This is why it's critical to build in opportunities for equal and meaningful participation from civil society, particularly organizations and communities from low- and middle-income countries that face additional barriers to participation. It also means creating pathways for public voices and affected communities to inform the Dialogue, recognizing that AI governance must be grounded in lived experience and public interest, not only expert or industry perspectives.

• Safe, secure and trustworthy AI
• AI capacity-building
• Interoperability of governance approaches
• Open-source software, open data and open AI models

The Global Partnership for Sustainable Development Data prioritizes these areas because they reflect the core conditions required for AI to support sustainable development based on what we've heard from our network of more than 700 actors across sectors working to advance data for sustainable development. Safe, secure, and trustworthy AI depend on the quality, representatativeness, and governance of the data that underpin AI systems as well as mechanisms for accountability, public participation, and oversight. AI capacity-building is critical because many governments, national statistical systems, civil society organizations, and local development partners lack the technical, institutional, and governance capacity to leverage AI for public good. Existing global inequalities in who shapes and benefits from AI will increase without targeted investments in capacity building. Interoperability of governance approaches is a priority given the increasing fragmentation of the global AI governance landscape. Greater coherence across national, regional, and international approaches can reduce duplication and friction and support effective cross-border collaboration. In this context, coordination with existing multilateral processes such as the Working Group on Data Governance mandated by the Global Digital Compact is essential to ensure alignment and avoid siloed approaches to AI and data governance. Open data with guardrails is an important enabler of inclusive AI, particularly for public-interest use cases. For many countries and institutions, meaningful participation in AI systems depends on access to high-quality representative, and responsibly-governed data. Expanding access supports innovation, transparency, and equitable outcomes. However, greater openness must be accompanied by strong data governance frameworks to ensure privacy, security, accountability, and appropriate safeguards. Responsible, equitable access to data enables AI to serve the public good.

A key cross-cutting issue is the foundational role of data governance in shaping AI outcomes. AI has the potential to revolutionize how public services are delivered, but this depends on access to diverse, high-quality data and the systems that enable its responsible use. Particularly, investments in public data and national data systems are key to making them AI-ready and to ensure that AI tools are not built on biased or incomplete data. AI should be understood as part of a continuum of existing investments in open data, digital transformation, and interoperability. Leveraging and strengthening these systems can expand access to more inclusive training data and address power imbalances and inequities. At the same time, new governance approaches will be needed to manage emerging data sources and uses. A second issue is the growing concentration of power across data, compute, and infrastructure. This affects competition, sovereignty, and the ability of public-interest actors to participate meaningfully in AI development and governance. Evening the playing field while enabling innovation requires coordination across policy domains and addressing gaps in capacity. Finally, implementation capacity remains a critical gap. Many countries lack the institutional, legal, and technical capacity to operationalize AI governance. Without greater focus on translating global frameworks into practice-particularly in low- and middle-income contexts-governance risks remaining aspirational rather than actionable.

Across our network, gaps in data access, stewardship, and coordination already constrain the ability of countries and development actors to engage meaningfully with AI. Many partners lack access to the diverse, high-quality data needed to develop or apply AI for sustainable development, whether because such data does not exist, is fragmented across systems, or remains inaccessible due to legal, technical, commercial, or capacity barriers. This limits the development of locally relevant, inclusive AI systems. At the same time, national partners are increasingly focused on data sovereignty, especially as AI systems rely on cross-border data flows and infrastructure that are often controlled by a small group of actors. This calls for balancing needs for openness and interoperability with the power to retain control over data and derive value domestically. These challenges are not new in the data for development community, but they are intensified by AI. Weak and fragmented data governance frameworks across the data lifecycle are being exposed and amplified by AI, increasing risks related to privacy, bias, exclusion, and misuse. Ongoing efforts, such as the CSTD Working Group on Data Governance, reflect growing recognition of these gaps but alignment with AI governance processes remains limited. At the same time, the surge in interest in building AI for good represents a significant opportunity to increase investments in data systems, digital public infrastructure, and open data, which are the building blocks of AI systems. Strengthening data governance enables more inclusive, locally-grounded AI applications to serve the public good.

The Dialogue plays a critical role in advancing international cooperation by moving from high-level principles toward coordinated, practical action. It provides a platform to align stakeholders while grounding global discussions in implementation realities. To do this, the Dialogue can: Develop a shared foundation of the building blocks of equitable AI governance: Foster convergence on core elements to provide a common reference point for national and regional governance approaches that are responsive to local contexts. Advance the development of standards for data for AI: Build consensus around the need for globally recognized standards in the use of data for AI—particularly around equitable access, transparency, and accountability—and lay out a pathway for their development. Bridge AI and data governance discussions: Align AI discussions with ongoing data governance processes to ensure coherence, recognizing that effective and equitable AI governance depends on the rules and norms that govern how data is collected, accessed, shared, used, and re-used. Coordinate the AI for good ecosystem: Reduce fragmentation by developing consensus on common challenges and opportunities to align efforts across governments, multilaterals, civil society, and the private sector, helping to focus resources on shared priorities for action. Center LMIC priorities and identify power asymmetries: Elevate capacity constraints, sovereignty concerns, participation gaps, while addressing concentration of power in data, compute, and infrastructure that shapes who benefits from AI. Advance interoperability across governance approaches: Promote greater coherence across national, regional, and international frameworks to support cross-border collaboration, including on data flows and safeguards. Connect global norms to implementation: Provide opportunities to translate principles into action by highlighting practical use cases, partnerships, and institutional models that enable responsible, inclusive AI systems to flourish in diverse contexts.

The Dialogue should build directly on the work of the CSTD Working Group on Data Governance, also mandated by the Global Digital Compact, which is already deliberating issues that are core to the AI for good community such as cross-border data flows, interoperability, value from data, and governance principles. This process reflects diverse perspectives, including from low- and middle-income countries, civil society, and the private sector, and is grounded in implementation realities faced by them. The Dialogue can add value by ensuring stronger alignment between AI governance and data governance, explicitly linking these discussions rather than allowing them to evolve in parallel, risking duplications and gaps. More broadly, there is a risk that AI governance and data governance processes across the UN system proceed on separate tracks despite their deep interdependence. The AI Dialogue can bridge these efforts, reinforcing coherence and ensuring that emerging AI governance approaches are grounded in existing areas of consensus on data governance. Finally, the Dialogue can elevate areas of convergence emerging from the CSTD process into broader AI governance discussions. By carrying forward shared priorities across stakeholders, it can help build a more coordinated and inclusive global approach to digital transformation.

We applaud the work of the Dialogue so far to seek broad input and develop opportunities for consultation with diverse stakeholders. At the 2026 Global Dialogue, we support making the deliberations and discussions of members as transparent and accessible as possible and seeking ongoing input at each stage of the process going forward. Significant barriers exist to ensuring stakeholders from around the world have opportunities to meaningfully and equally participate in global, multilateral processes. That's why we also support using the first Global Dialogue to showcase existing public voices by providing exhibition space, side event opportunities, space in the speakers list and breakout sessions for representatives from civil society and citizens' groups.

AI is reshaping societies far beyond the places where models are developed or deployed most visibly. Its effects are felt in healthcare, education, employment, public services, labor markets, media ecosystems, and the environment—including by people who are not digitally connected or direct users of AI systems. Yet global AI governance fora continue to overrepresent governments, large technology companies, and highly resourced institutions, while underrepresenting affected communities, especially in low- and middle-income countries. This includes youth and older people, Indigenous communities, informal workers, data labeling and content moderation workers, civil society organizations, and communities whose data is used to train or inform AI systems without meaningful participation or benefit. It also includes national and local public institutions—such as statistical offices and regulators—that are responsible for governing data and managing the downstream impacts of AI adoption. Including these voices requires moving beyond highly formalized, expert-led, and primarily digital consultation models. Community organizations, local institutions, and trusted intermediaries are essential for enabling "last-mile" participation and ensuring engagement reflects lived realities rather than only technical or policy expertise. Distributed dialogue models—such as local assemblies, sector-based consultations, and civil society-led convenings—can help surface perspectives that would otherwise be excluded.

In addition to formal panels and expert sessions, the AI Dialogue should create space for direct exchange between policymakers, technical experts, and affected communities. We support models such as live link-ups with citizen assemblies and community dialogues happening in parallel around the world, allowing participants to bring questions and perspectives directly into the Dialogue. This helps connect global governance discussions to everyday concerns and strengthens public trust and legitimacy.

Several existing models offer practical lessons for effective AI governance by linking high-level principles to implementation. The OECD AI Principles and the UNESCO Recommendation on the Ethics of AI have helped establish shared global norms, but their value has come from accompanying tools for implementation such as UNESCO's Readiness Assessment Methodology, which helps countries assess institutional, legal, and capacity gaps before adopting AI systems. At the national level, Brazil's data protection authority and India's Digital Public Infrastructure approach show the importance of building governance around data systems, not only AI applications. Investment in national data systems, stronger data governance, public-interest safeguards, and institutional coordination across regulators, statistical offices, and digital agencies are essential preconditions for trustworthy AI. In the public-interest data space, trusted data-sharing models also provide useful examples. Data collaboratives and data stewardship approaches, such as public-private partnerships enabling the use of mobile positioning data for official statistics, demonstrate how privately-held data can be used for development outcomes when supported by clear governance frameworks, legal safeguards, and accountability mechanisms. Participatory governance models are equally important. Taiwan's vTaiwan platform and citizens' assemblies on AI in countries such as France and the UK show how governments can incorporate public input into technology governance decisions, helping ensure legitimacy and trust. These approaches are particularly valuable for addressing questions of social impact, labor transitions, and public service delivery that cannot be resolved through technical standards alone.