Digital Action

A successful first Global Dialogue on AI Governance would be one that breaks down existing power asymmetries and creates equitable spaces for engagement for those who have been historically marginalized in a multilateral setting. Success should be measured by whether the process meaningfully shifts whose knowledge, lived experience, and priorities shape the governance agenda. The Dialogue should explore decentralized participation models that can reach grassroots organizations, Indigenous communities, and those living under AI-driven harm who rarely participate in multilateral processes. From Digital Action's perspective, other outcomes would constitute genuine success. First, the Dialogue should center the voices of countries and communities that bear the compounded burdens of AI's extractive logics, including the environmental costs of data center expansion, the depletion of water and energy resources in already-stressed regions, and the displacement of local economies in the name of digital infrastructure. These communities have had the least say in how AI is governed, yet absorb its heaviest costs. Their priorities should be central, not a regional footnote. Second, the Dialogue should establish a clear normative commitment to grounding AI governance in international human rights law. Principles of transparency, accountability, and non-discrimination are binding obligations and not optional features. The Dialogue should close the gap between invoking these principles and enforcing them. Third, the Dialogue should address AI capacity gaps in their full dimension, technical, institutional, and participatory. This means sustaining locally-driven regulatory capacity, research infrastructure, and civil society expertise, enabling Global South communities to evaluate, contest, and govern AI systems on their own terms. A Dialogue that achieves these outcomes would build the legitimacy and trust necessary for deeper engagement in 2027 and beyond. One that falls short risks becoming another high-level forum that speaks about inclusion while structurally excluding those most affected.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

The protection and promotion of human rights is the normative foundation. It establishes the non-negotiable floor from which all governance frameworks should be built, including the rights to remedy, non-discrimination, privacy, and free expression. Without this anchor, governance risks becoming a technical exercise disconnected from the lives of those most affected. The social, economic, ethical, cultural, linguistic, and technical implications of AI constitute the diagnostic layer. This is where the actual costs of AI development become visible: the environmental burden of data center expansion on water- and energy-stressed communities, the erasure of minority languages from training datasets, the displacement of local labor markets, the exploitation of low-wage workers in the Global South to perform invisible labor of data annotation and content moderation that makes AI system function, and the embedding of racial and gender bias into automated decision-making. Transparency, accountability, and human oversight are only meaningful if the capacity to exercise them is distributed. Oversight frameworks that concentrate technical and institutional power in a small number of actors, without relying on the ability of Global South countries and communities to audit, contest, and regulate AI systems on their own terms, reproduce dependency rather than close the governance gap. There's an opportunity to include local and community-based AI governance capacity: the expertise, infrastructure, and institutional standing to make transparency and accountability real where it matters the most. Interoperability of governance approaches is where these three priorities meet their hardest test. It should create space for a plurality of legal traditions, development contexts, and social values, rather than harmonizing frameworks around standards of the most powerful actors. Together, these priorities ask not only how AI should be governed, but by whom, for whom, and from where.

The most significant omission in the current thematic framework is the environmental and climate dimension of AI. This is not a peripheral concern; it is a structural feature of how AI systems are built and sustained, with consequences that fall disproportionately on communities in the Global Majority. AI is not a cloud. It is a physical infrastructure of fossil-fueled data centers, undersea cables, and semiconductor supply chains that consumes vast quantities of water, minerals, and energy. These facilities are increasingly sited in regions already under climate stress, where communities have the least regulatory capacity to contest their presence or demand accountability for their impacts. These same systems are being deployed in extractive industries accelerating ecological harm and deployed without governance frameworks that protect indigenous and traditional knowledge systems or give communities the right to determine how, or whether, their knowledge is used. The absence of binding disclosure requirements for the environmental footprint of AI systems represents a concrete governance gap that the Dialogue should name and address. A second critical omission is the weaponization and militarization of AI. Civilian AI systems are routinely repurposed for surveillance, targeted harm, and conflict. The Dialogue has an opportunity to address dual-use risks explicitly, ensuring that the non-military mandate does not become a shield against accountability for AI systems deployed to cause harm. Both omissions point to a structural problem: AI governance and broader planetary and security governance are running on parallel tracks within the UN system. The Dialogue has a unique opportunity and responsibility to bridge them, establishing concrete linkages with multilateral environmental agreements and existing accountability frameworks. Governance conducted in silos cannot address challenges that do not respect institutional boundaries.

As a cross-regional organization working across Latin America, Africa, and the South West Asia and North Africa (SWANA) region, we observe a consistent pattern: communities in the Global Majority are subject to AI systems they did not design, governed by frameworks they did not shape, yet have had the least say in the governance decisions that will determine whether AI's opportunities reach them or bypass them entirely. In Latin America, AI deployment is outpacing governance across multiple fronts: facial recognition technologies operating in public spaces with minimal oversight, content moderation systems with inadequate capacity in local languages, and data center expansion imposing disproportionate environmental burdens on water- and energy-stressed communities. Across Africa, national AI strategies are proliferating rapidly, outpacing AI development, yet the majority lack rights-based frameworks commensurate with the risks they govern. The AU Continental AI Strategy provides a regional reference point, but implementation is uneven, and regulators, innovators and civil society still grapple with the capacity to meaningfully engage in national policy processes that are severely constrained by resource limitations, technical complexity, and deliberate opacity in AI development and deployment. In addition, AI technologies are influenced by foreign investments, and the state has leveraged AI-enabled tools to strengthen its capacity for citizen monitoring and control. In the SWANA region, AI governance gaps are compounded by shrinking civic space and governments that are influenced by foreign investments. National AI strategies are emerging in contexts where AI technologies, including facial recognition, predictive policing, and content moderation systems, are deployed in ways that reinforce state control and exacerbate existing inequalities. Civil society participation in governance processes is structurally blocked in many contexts, creating accountability vacuums with no effective remedy mechanisms. Across all three regions, the governance gap is not only technical or institutional; it is political. Frameworks designed elsewhere are applied universally, without regard for local legal traditions, social contexts, or the communities whose knowledge, agency, and self-determination should be at the center of any legitimate governance process.

The AI Dialogue's most distinctive contribution to international cooperation is structural: it is the only governance forum with universal membership, civil society inclusion, and a mandate to address AI governance as a matter of global public interest rather than regional or sectoral priority. That structural position carries responsibilities that go beyond convening; it should actively shape what international cooperation means in practice. First, the Dialogue should establish that international cooperation on AI governance is not synonymous with convergence. Meaningful cooperation requires creating the conditions for a plurality of governance approaches to coexist and inform one another, approaches rooted in different legal traditions, development contexts, and social values. The Dialogue's role is to build bridges between these approaches without flattening their differences or privileging any single framework as the default. Second, the Dialogue can advance cooperation by rebalancing whose knowledge and experience informs global standard-setting. International governance processes have historically reflected the priorities of the most technologically advanced actors. The Dialogue has a unique opportunity to systematically incorporate evidence, expertise, and governance innovations from the Global Majority into the shared knowledge base from which international norms are built, not only as regional inputs, but as constitutive contributions. Third, the Dialogue should deliberately use its cumulative architecture. Building a normative trajectory and establishing shared understandings that progressively create pressure toward binding commitments in other forums. This requires the Dialogue to be explicit about the direction of travel: toward governance frameworks grounded in human rights, environmental accountability, and genuine democratic participation. Cooperation, in this sense, is not an end state. It is a practice of building shared legitimacy from the ground up.

Beyond intergovernmental frameworks, the Dialogue should actively engage with civil society knowledge networks that have developed substantive expertise in rights-based AI governance across the Global Majority. The Global Coalition for Tech Justice, convened by Digital Action and comprising over 250 member organizations across 55 countries, has developed concrete proposals (https://digitalaction.co/projects/building-the-ai-commons-join-us-in-democratizing-ais-future/) — including democratic oversight mechanisms, capacity-building infrastructure, and community-centered design frameworks — that represent actionable civil society input the Dialogue should draw on. Regional civil society networks across Africa, Latin America, and the SWANA region similarly carry granular, contextually grounded knowledge that intergovernmental processes could rely on. The Dialogue's added value lies precisely in what these other initiatives cannot provide: a universal, multistakeholder forum with the legitimacy to build shared understandings across the full spectrum of governance approaches. Its unique contribution is not to produce another framework, but to create the connective tissue between existing ones, ensuring that norms developed in regional or thematic contexts can inform, challenge, and strengthen global governance without being subordinated to the preferences of the most powerful actors.

The current architecture of global AI governance, concentrated in a small number of cities, conducted primarily in English, and structured around formal delegation models, systematically excludes the Global Majority: the communities, movements, and knowledge systems whose participation is not peripheral to legitimate AI governance but constitutive of it. We call for a polycentric (Scholte, 2017) participation model that distributes the Dialogue's deliberative weight across multiple nodes, regional, national, and community-level, rather than concentrating it in a single annual event in Geneva. Regional pre-dialogues should function as genuine agenda-setting mechanisms, not only consultations whose outputs are noted and set aside. Their findings, framed in the voices of the communities that produced them, should be substantively reflected in the Geneva proceedings and in the co-chairs' summaries. Stakeholder contributions should be reconceived beyond the standard typology of governments, private sector, and civil society. Affected communities, those living under AI-driven surveillance, displacement, or environmental burden, should have direct participation mechanisms to present evidence and testimony. Grassroots movements, Indigenous knowledge holders, and youth networks carry expertise that formal delegation models routinely miss. The Dialogue should also focus on participation infrastructure across languages and formats. Real inclusion cannot be contingent on the ability to travel to Geneva or engage in real-time in English. Documentation, deliberations, and outcomes should be accessible across languages and formats from the outset, not as an afterthought. We believe the Dialogue has an opportunity to model a different vision of international governance; one where digital ecosystems embody our different realities, advance social justice, and sustain human rights. That vision begins with how the Dialogue itself is structured: not as a space where the Global Majority is consulted, but where it leads.

Underrepresentation in global AI governance is not accidental; it is structural. Addressing it requires identifying the mechanisms that produce exclusion, not simply enumerating the groups excluded. The first axis is linguistic. Global AI governance deliberation occurs overwhelmingly in a handful of languages, rendering structurally inaudible the epistemic traditions, legal concepts, and governance innovations of communities whose knowledge exists in other languages. Indigenous peoples, linguistic minorities, and communities across the Global Majority are not merely underrepresented; their ways of knowing are systemically filtered out before they reach the table. Inclusion here requires genuine multilingual infrastructure: not translation as an afterthought, but deliberation designed from the outset to be legible across languages and knowledge systems. The second axis is institutional. Formal participation in UN processes requires organizational accreditation, resources to travel, and familiarity with procedural norms that community-based organizations, grassroots movements, and informal networks rarely possess. Youth networks, feminist collectives, informal worker organizations, and community-led initiatives carry governance knowledge that the current model systematically misses. Trusted intermediary networks and civil society organizations with deep community roots can help bridge this gap, but this requires treating them as governance actors in their own right, with real agenda power and dedicated resourcing. The third axis is political. In many Global Majority contexts, civic space is shrinking. Digital rights defenders, environmental advocates, and community organizers face legal, institutional, and physical risks that make formal international participation dangerous or impossible. The Dialogue should establish protected, confidential participation channels for those operating in restricted civic space environments, recognizing that silence in these contexts is not the absence of perspective, but evidence of suppression. Inclusion that addresses only the first axis while ignoring the second and third is inclusion in name only.

The first Global Dialogue on AI Governance is an opportunity to model a different kind of international conversation. One that is generative, accessible, and genuinely open to the unexpected. We offer four format innovations that could make that vision concrete. Community testimony sessions would create a structured, guaranteed time for affected communities to present evidence and lived experience directly to governments and stakeholders. Their own voices and languages. Governments and other actors would have response obligations, transforming testimony from performance into genuine dialogue with accountability. Living documentation would make the Dialogue's deliberations accessible in real time through multilingual, user-friendly, publicly available records that allow those outside the room to follow, respond, and contribute during the Dialogue itself. Documentation designed for accessibility from the outset, not archived afterward, transforms the Dialogue from a closed event into an open process. This is also a governance innovation: it creates a shared, contestable record that communities can use to hold participants accountable to what was said. Cross-regional peer exchange sessions would create a structured space for Global South practitioners, like policymakers, civil society actors, and researchers, to share governance innovations with each other and with the broader Dialogue. Centering south-south knowledge flows rather than north-south transfer recognizes that some of the most creative AI governance responses are emerging precisely where resources are most constrained. Finally, we propose dedicated futures and imagination sessions, spaces for speculative, creative engagement with what AI governance could look like beyond current paradigms – including policy hackathons and regulatory sandbox pilots. Governance that cannot imagine otherwise cannot build otherwise. These sessions, drawing on diverse cultural and epistemic traditions, would expand the horizon of what the Dialogue considers possible, and remind participants that the goal is not just better rules, but better worlds.

Effective AI governance approaches share a common feature: they treat governance as a political and social practice that requires redistributed power and diverse knowledge, not merely a technical problem to be solved. We identify other approaches that have demonstrated concrete value. Rights-based assessment frameworks apply international human rights law as the primary analytical grid for evaluating AI strategies and systems, examining alignment with principles of non-discrimination, privacy, transparency, accountability, and the right to remedy. This approach grounds governance in binding legal obligations rather than voluntary commitments, and provides affected communities with contestable standards against which to measure compliance. Civil society knowledge infrastructure, systematic documentation of lived experience and cross-regional research, is in itself a governance input, ensuring the Dialogue has access to knowledge that institutional processes alone cannot produce. Community-centered design and oversight include affected communities in AI development, testing, and ongoing monitoring as co-architects of the systems and rules that govern them. This approach directly addresses capacity gaps by building local expertise and decision-making power where it is most needed. Feminist and intersectional impact assessment embeds gender, racial, and environmental lenses into AI governance evaluation from the outset. This approach surfaces harms that technical safety frameworks routinely miss and ensures that trustworthiness is measured against the experiences of the most marginalized, not the most privileged. South-south governance exchange creates horizontal knowledge flows between Global Majority regions, sharing regulatory innovations, governance experiences, and civil society strategies across contexts that face similar challenges. This approach builds genuine interoperability grounded in shared realities rather than imported frameworks. What distinguishes effective governance is the accountability pathways built alongside the principles declared and starts from the beginning of the technology's lifecycle. Implementation requires independent monitoring, accessible remedy mechanisms, and enforcement that goes beyond voluntary compliance, ensuring that when international human rights law is invoked, communities can actually hold systems and institutions to it.