Aapti Institute

The success of the first Global Dialogue on AI Governance will depend crucially on the breadth of its participation. Genuine Global South representation is not merely a procedural nicety but a substantive necessity. The AI value chain is geographically distributed, spanning labour, data, hardware, and infrastructure across dozens of countries, and this distribution has produced governance challenges that are both shared and geography-specific. No single nation can address these alone. Effectively governing the global AI value chain requires first decoding the different flows of resources that sustain it, and then building the collaborative frameworks to regulate them. This is made harder by a structural imbalance. AI technology, compute infrastructure, and corporate power are overwhelmingly concentrated in the Global North, which skews the power dynamics of any governance effort. Countries outside these power centres have limited leverage to enforce compliance from multinational technology companies operating within their borders. This asymmetry must be named and addressed directly, not treated as background context. Further, AI capacity gaps in developing countries are not simply a development problem but a governance problem. Nations without the technical expertise or infrastructure to participate meaningfully in an AI-first world will find their futures shaped by decisions made elsewhere. Closing these gaps must be a core output of the Dialogue, not an afterthought. To this end, the equitable distribution of responsibility through regulation and policy is key. Policy must account for the flow of power and impact through the AI value chain and assign proportionate responsibility to ensure progress and development in Global South countries is not hindered. Additionally, the responsible AI governance landscape is already fragmented across multiple overlapping frameworks. For the Dialogue to provide real added value, it must work to bring coherence to existing initiatives. Its universal mandate is its greatest asset.

AI capacity-building;Protection and promotion of human rights;Transparency, accountability, and human oversight;Open-source software, open data and open AI models

1. AI capacity-building: The technology, compute infrastructure, and corporate power that power AI are overwhelmingly concentrated in the Global North. This structural imbalance means that countries in the Global South are subject to exploitation for labour and remain largely excluded from its governance, given the lack of regulatory understanding and control to enforce governance. Closing AI capacity gaps in developing countries is, therefore, not a development priority alone but a governance imperative. Without the technical expertise and regulatory institutions to participate meaningfully, these nations will find their futures shaped by decisions made elsewhere. 2. Open-source: Meaningful open-source models, open data, and shared infrastructure are among the most practical tools available to counteract the concentration of resources, enabling broader participation in both the development and the oversight of AI systems. For countries and communities without access to frontier proprietary models, openness can be a functional necessity for meaningful engagement in an AI-first world. 3. Transparency, accountability, and human oversight: Governing the global AI value chain requires first being able to see it clearly. Without transparency into how AI systems are built, trained, and deployed, accountability becomes impossible and human oversight becomes nominal. This is particularly urgent given the difficulty that countries outside major AI power centres face in enforcing compliance from multinational technology companies. Transparency obligations create the evidentiary foundation on which any meaningful governance framework must rest. 4. Protection and promotion of human rights: AI governance previously has not been anchored in human rights risks, and has largely become a technical exercise that optimises systems. Context-specific harms, labour exploitation in AI supply chains, AI-enabled surveillance, and algorithmic discrimination all have direct human rights dimensions. These cannot be addressed through voluntary corporate commitments alone and require the kind of binding, universal framework that only a multilateral forum like this Dialogue can begin to build.

1. Environmental sustainability: The compute infrastructure underpinning AI systems consumes extraordinary and rapidly growing quantities of energy and water, with data centres disproportionately sited in regions that bear the environmental costs without sharing proportionately in the economic benefits. As AI adoption accelerates, the carbon footprint of training and running large models will become a significant variable in global climate commitments. AI governance cannot be conducted in isolation from climate governance, and the Dialogue must address sustainability not as a peripheral concern but as a condition of responsible AI development. 2. Labour in the AI value chain. Data workers, content moderators, and annotators who make AI systems function represent a largely invisible workforce, concentrated in the Global South, operating under precarious conditions with minimal regulatory protection. AI governance that only addresses AI deployment without addressing the conditions of AI production remains incomplete. The Dialogue should work toward establishing baseline labour standards for AI supply chains and creating formal mechanisms for these workers to participate in governance processes that directly affect them. 3. Gendered dimension of AI: Both as a structural bias embedded in AI systems and as a dimension of who governs them. Gender discrimination reproduced through algorithmic systems, the particular vulnerabilities of women to AI-enabled harms such as deepfake abuse and surveillance, and the systematic exclusion of women's rights perspectives from technical and policy decision-making all require explicit treatment. A gendered lens must be applied across every thematic area of the Dialogue rather than confined to a standalone discussion that the rest of the agenda treats as supplementary.

1. Protection and Promotion of Human Rights India's AI governance gaps have direct and specific human rights implications, particularly for labour. The country is a significant hub for data annotation, content moderation, and AI training work that underpins global AI systems, yet this workforce operates with minimal regulatory protection, limited collective bargaining rights, and almost no representation in governance discussions. Further, technology-facilitated gender-based violence is a widespread yet under-acknowledged crisis in India, and AI has significantly expanded both the scale and sophistication of this abuse. Individuals from marginalised groups are being targeted through doxing, stalking, morphing, and the non-consensual distribution of intimate images. Existing legal frameworks are not only insufficient but also frequently inaccessible to survivors, leaving harm unaddressed and perpetrators unaccountable. This inaccessibility is also magnified because of the limited intermediary liabilities taken on by multinational tech giants operating in India. 2. Environment: India has been experiencing a rapid growth in data centre investment, driven by both domestic AI ambition and the appetite of global technology companies for lower-cost infrastructure. This expansion carries significant environmental costs in a country already under severe climate stress, raising urgent questions about energy sources, water consumption in water-scarce regions, and whether the communities hosting this infrastructure share in its benefits. 3. AI Capacity-Building: India possesses significant technical capacity, an ambitious national AI mission, and a thriving startup ecosystem. Yet, large language models and generative AI systems penetrating Indian markets, public services, and workplaces are built by a handful of multinational corporations that neither capture the realities of Indian users nor provide meaningful input into decision-making processes that affect them. India's ability to shape the AI systems its citizens interact with, ensure they reflect Indian contexts, and values, and enforce accountability when they cause harm, is constrained by where foundational AI infrastructure and intellectual property are concentrated.

The AI Dialogue occupies a unique and necessary position in the international governance landscape. Its most valuable role is therefore not to duplicate what regional frameworks, voluntary coalitions, or industry-led initiatives already do, but to function as the mediator across them. Currently, AI governance exists in a fragmented governance landscape, with multiple frameworks, principles and declarations, which run parallel without coherence. This incoherence creates regulatory arbitrage opportunities that well-resourced actors will exploit and that smaller, less-resourced countries will be least equipped to navigate. The Dialogue can address this by working actively toward interoperability between existing frameworks, identifying where they converge, surfacing where they conflict, and building the shared vocabulary that makes coordination across them possible. Beyond coherence, the Dialogue has a critical role to play in rebalancing who shapes the global governance agenda. The concentration of AI power and technology means that governance norms have largely been set by the same actors who benefit most from minimal constraints. By giving every country a genuine seat at the table, the Dialogue creates the conditions for governance that reflect the full range of interests at stake, including those of the communities most affected by AI systems, who have had the least say in how they are built or regulated.

The Dialogue should build on the work of the UN Commission for Science and Technology for Development (CSTD) Working Group on Data Governance. Data governance is a vital yet underemphasised tenet in current discussions, playing a central role in enabling responsible AI development, cross-border data flows and innovation landscapes. This engagement will situate AI governance within the broader political economy of data, including data access, sharing, stewardship and value creation. Explicitly linking AI-related recommendations with discussions on data governance at the UN level will prevent the isolation of AI from the infrastructures and institutions that determine who controls and benefits. The UN CSTD Working Group will provide a foundation to further dialogue on effective and equitable data governance frameworks. There is also value in engaging with the OHCHR to institutionalise the ongoing process of a binding legal instrument for transnational cooperation and human rights, particularly to ensure accountability in AI-driven business models. Guiding principles for Business and Human Rights for various stakeholders provide assessment tools, support capacity building and promote shared mechanisms that can be built upon and adapted for the Dialogue. The Global South Alliance (GSA), comprising 26 member organisations with expertise in digital rights is another key stakeholder network to engage with through the Dialogue. The GSA's active engagement with the UN Global Digital Compact process offers valuable perspectives from the Global South.

While national government representation brings the sovereign authority necessary to translate governance commitments into binding national frameworks and must remain the primary decision-making constituency, their deliberations should be systematically informed by other categories of actors, each requiring a distinct and meaningful mode of participation. Civil society organisations, particularly those rooted in affected communities in the Global South, should have formal agenda-setting rights, not merely observer status. This means structured opportunities to introduce evidence, challenge framing, and propose priorities. Further, academic and scientific contributors, operating through the Independent International Scientific Panel, should feed directly into each session with findings that are accessible to non-specialist delegates and translated into concrete governance implications rather than presented as standalone reports. One way to incorporate such perspectives can be intersessional working groups organised around specific thematic areas, meeting between annual sessions, which would allow deeper technical work to happen continuously. Dedicated spaces for data workers, women's organisations, and youth representatives should be built into the formal programme architecture of the Global Dialogue. A useful model is the United Nations Framework Convention on Climate Change (UNFCCC) governance architecture, where intersessional bodies such as the Subsidiary Body for Scientific and Technological Advice meet regularly between COP sessions to advance technical work, feed findings into the main negotiating process, and maintain continuity across the annual cycle.

Global discussions on AI governance have remained dominated by a narrow set of voices, mainly technologists, policymakers, and corporate representatives. However, there remain other actors in the AI value chain who actively participate in its building yet remain effectively invisible in governance conversations: 1. Data workers, content moderators, annotators, and prompt engineers, the majority of whom are based in the Global South, perform the foundational work of training and refining AI models under conditions that are frequently precarious, poorly compensated, and psychologically harmful. Their exclusion is not incidental but structural. Any serious effort to govern AI must name this labour, regulate the conditions under which it happens, and create formal mechanisms for these workers to participate in decisions that directly govern their working lives. The AI Dialogue, with its universal mandate and multistakeholder architecture, is one of the few forums positioned to make that inclusion real. 2. Users of AI like farmers navigating AI-driven agricultural pricing systems, patients in under-resourced healthcare systems encountering algorithmic triage, students in under-funded schools using AI tools shaped by curricula from elsewhere, and residents in communities hosting data centres bearing environmental costs they did not choose. These are the people whose lived experiences should be forming the evidentiary base of global AI governance, and they are almost entirely absent from it. Bringing them in requires consultation processes that are designed from the outset to receive experiential knowledge, not just technical expertise.