The South Asians for Digital Rights (SADR) Coalition

Roadmap for better AI Governance: The success of the Global Dialogue will hinge on developing a 5-year strategy for AI governance processes and regulations. As the conversation around AI evolves, it is pertinent to establish set tenets that serve as guidelines for the various stakeholders focusing on actionable principles and regulatory frameworks. Responding to prevailing global order: The conversation needs to be located and respond to the erosion of digital rights and the impact of AI supply chains on the global south. Solutions and outcomes need to address these threats as well as issues arising in the global order — on the environment, democratic backsliding, and labour — and create pathways for underrepresented stakeholders to meaningfully participate. Building a layer of transparency: Clarity on how non-state stakeholders such as civil society, businesses, the technical community, academia, and investors will participate in the Dialogue, as well as openness regarding which initiatives, partnerships, or case studies will be elevated and showcased. Fostering meaningful multistakeholderism: Multistakeholder participation is only meaningful if it is structurally embedded. This means transparent selection criteria that distinguish grassroots and community-based organisations from well-resourced intermediaries, dedicated funding for underrepresented actors to participate, and session formats that allow civil society and affected communities to shape outputs rather than simply be present in the room. An explicit Co-Chairs' Summary: The Co-Chairs' Summary should be explicit in where convergence exists and where it does not. The risk in any process like this is that contested questions are responded to with ambiguous goals and objectives. Non-state actors need a record they can point to to advance their positions in subsequent processes.

• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

Our selection prioritizes issues where structural exclusion meets urgent harm. 1. AI capacity-building: The ability to develop, evaluate, and regulate AI systems is concentrated in companies and jurisdictions in the Global North. South Asian communities are structurally excluded from AI governance due to a lack of regulatory infrastructure, technical expertise, and a civil society knowledge base. Capacity-building means addressing the structural gap in AI transformation with long-term goals and priorities that are context-specific, not just short-term funding and training programs. 2. Social, economic, ethical, cultural, linguistic, and technical implications: Historical codification and biases, combined with new forms of labour exploitation, make this a pertinent issue. Addressing the most urgent harms, including algorithmic management of gig and data workers, AI-amplified disinformation, and the cultural and linguistic marginalisation of South Asian communities in AI systems designed with little or no community involvement, is creating an ecosystem of invisibilisation. System vulnerabilities and social biases, with no legal recourse or accountability mechanisms, put users at risk. Conversations need to push for finding ways to co-create systems and address historic biases. 3. Protection and promotion of human rights: AI systems in our region are used to conduct mass surveillance, deny welfare and identity services, and automate discrimination against marginalised communities in our region. There is a need to go beyond voluntary corporate commitments and push for a binding, universal framework that only a multilateral forum like this Dialogue can begin to build. The human rights framework is the only globally agreed baseline that predates any particular AI governance regime and must anchor the Dialogue's work. 4. Transparency, accountability, and human oversight: Communities cannot challenge systems they cannot see. Meaningful oversight means that affected communities, not only regulators or developers, have access to the information needed to understand and contest AI-driven decisions that affect their lives. Transparency obligations create the evidentiary foundation on which any meaningful governance framework must rest.

1. Algorithmic management & Data Work: South Asia is a major outsourcing hub, with India among the largest sites for data work. Asymmetrical power dynamics exist between Global North tech companies and South Asian data workers. Due to the transnational dispersal of these workers, there are severe constraints on collective bargaining and accountability. The Dialogue can advance conversations on social security, occupational safety, the subject of labour to technology, and transnational business arrangements. 2. Environmental Costs of AI: Resource extraction for hardware, e-waste disposal and data centres disproportionately affect South Asia. The benefits of both tech advancement and economic gains are skewed in favour of the Global North. How value is redistributed across the supply chain must be addressed for an equitable and sustainable growth trajectory. 3. AI in conflict and surveillance contexts: AI-enabled surveillance is being undertaken during a time of political instability, often without adequate information about the data localisation and retention. Although undertaken on a large scale and with highly sensitive information, there is little publicly available information on the safeguards governing its use. The use of AI in warfare and conflict is another critical area that needs attention from an international humanitarian law and human rights law lens. Critical assessment needs to be done that moves beyond the rhetoric of public safety and enforces a clear and specific legal framework. 4. Data sovereignty and intellectual property: South Asian communities have little control over how their data is extracted, stored, or monetised. AI companies take both data and custody from owners, with no equitable return flow back to those who generated it. Meanwhile, AI models trained on local knowledge and creative outputs risk enclosure through Global North IP regimes. The Dialogue must clarify how data governance and IP frameworks can protect, not further appropriate, regional knowledge assets.

The most significant governance gaps in our region cluster around four areas. 1. Algorithmic management & Data Work: With a lack of meaningful regulatory frameworks for AI-driven labour management, gig workers and data workers across South Asia are subject to precarious working conditions. Algorithmic wage-setting, performance surveillance, and instant termination with no appeal mechanism need to be addressed at the state and regional levels due to the dispersed nature of workers. India's nascent data protection framework and the absence of equivalent legislation in Bangladesh, Pakistan, and Sri Lanka mean these workers have no legal recourse against the systems that govern their working conditions. 2. Environmental costs of AI infrastructure: Data centres induce further water and heat stress in South Asia. No current AI governance framework addresses these harms at source; the exacerbation of climate change-related events and the intersection of AI and ecological impacts need to be addressed. 3. Public systems and surveillance: With AI systems being deployed in the region, for instance, in welfare distribution, law enforcement, warfare, and financial services; they are predominantly developed elsewhere, optimised for different contexts, and deployed without local auditing capacity. There is a need to create metrics and infrastructure for social auditing, allowing for active community participation and a redressal mechanism for affected communities. 4. Data sovereignty and intellectual property: Because no regional framework protects data sovereignty, South Asian communities cannot challenge when their data is monetised without their consent or when their local knowledge, from traditional medicine to folk art, is scraped to train commercial AI models. Communities watch their digital and cultural assets generate value for Global North companies while receiving nothing in return. The opportunity the Dialogue presents is to establish that the communities most affected have a right to shape the frameworks that govern the systems affecting their lives.

1. Interoperability between different governance models: Consensus building: As a universal forum, Global Dialogue can help envisage reference points, allowing countries to coordinate across different regulatory traditions without any one framework becoming the baseline, recognising nuances of context, challenges, and key opportunities. Mapping the current landscape: Countries, especially in South Asia, lack the capacity to engage across all forums, leaving them absent from the conversations where norms are established. The Dialogue plays a critical role in mapping existing work, identifying gaps, and tasking an existing body, such as the The Internet Governance Forum (IGF) secretariat, with maintaining a public record of commitments and progress between sessions. Move towards an actionable objective: The Dialogue needs to act as an accountable mechanism and push for delivery. This can play a key role, especially for countries still waiting for the GDC's capacity-building and funding to operationalise. 2. Stronger play for CSOs' knowledge sharing: Democratising the conversation: Foster inclusive discussion involving different sectors, countries, and diverse perspectives Facilitating Multistakeholderism: The Dialogue's structural advantage is meaningful only if its processes respond to the perspectives of those with the least power and the greatest exposure to AI-driven harm. Thus, civil society, affected communities, and technical experts need to be platformed to shape outputs and to embed themselves in the AI discourse going forward. 3. Address some key questions surrounding: Emerging forms of Labour: How are data workers and gig labourers powering AI systems across Asia? What are the pathways to build cross-border solidarity? Land and Resources: Who bears the costs of data centres, from impacts on land, water, and electricity, to responsibility for decommissioning when infrastructure outlives its utility?

Building on Previous Initiatives: Existing multilateral frameworks and processes, such as the Global Digital Compact and its implementation architecture, particularly the Data Governance Working Group under CSTD and the Global Fund on AI. This will ensure that AI development and deployment are grounded in principles of social justice, inclusion, and fairness and build on conversations focusing on data stewardship, access, interoperability, and the mitigation of structural inequalities embedded in data ecosystems. The Dialogue's added value is to hold those commitments accountable and push for delivery, especially for countries still waiting for the capacity-building and financing promises of the GDC to translate into anything concrete. Centring Human Rights: Working in collaboration with the OHCHR to build on the process for a binding legal instrument for transnational cooperation and human rights. This comes at a time when notions of AI safety and harms are not prioritised in conversations: for the workers, users and ways they are being used against marginalised communities. The Dialogue can play a key role in ensuring accountability in AI-driven business models. There is a need to align tech innovation and human rights. Regional and Decentralised Systems: IGF—with National and Regional Levels—provides infrastructure for inclusive, multistakeholder AI governance. Regional IGFs, in particular, provide a mechanism for platforming locally grounded perspectives and feeding them into global processes. The Dialogue should build on these mechanisms rather than building new systems. Meaningful Multistakeholderism: The IGF and its Asia-Pacific regional network offer the most developed model for inclusive multistakeholder governance in our region. The NETmundial process and its multistakeholder guidelines provide a proven framework for ensuring civil society, affected communities, and technical experts from South Asia have a genuine role in shaping outputs. The Dialogue's added value is to operationalize these models.

The discourse surrounding AI needs to become more participatory and inclusive, moving beyond logics of technocracy to foster community-centred values rooted in context and to gain cognisance of the planetary boundaries. Key stakeholders should include: Regional Coalitions: help bring consolidated perspectives, emphasise shared experiences, Civil society perspectives: labour organisers, environmental justice advocates, and digital rights researchers to confront the hidden costs of AI's supply chain and build a shared advocacy agenda for accountability. The format and structure of the Global Dialogue should also ensure that stakeholders have the information and resources needed to participate meaningfully. Creating an accessible online platform: Regularly updated information, making materials available in the widest possible range of languages, sharing background documents and guiding questions with adequate advance, enabling transparent written feedback, and providing support for travel, connectivity, visa processes, low-bandwidth participation, and recorded sessions with synthesised summaries made publicly available. Enabling cross-sector collaboration: Working groups comprising government, civil society, technical, and private-sector representatives should be tasked with producing synthesised recommendations on a specific implementation challenge by the end of the event. Addressing key issues: Conversations should lead to tangible outputs and create genuine interdependence between stakeholder groups rather than siloed contributions. Examples include how to build a multilingual AI evaluation benchmark for low-resource languages or how to design a redress mechanism for cross-border AI-driven hiring discrimination. Building pathways to take the conversations forward: Establishing a lightweight intersessional mechanism, potentially anchored in the IGF, to track whether commitments made were acted upon before the second session. Without this, the first Dialogue risks being a productive conversation with no accountability thread connecting it to the second.

AI governance is yet to address the communities and regions which are bearing the brunt of AI development in the form of labour, data and natural resource extraction. In South Asia, this includes: 1. Gig and data workers who face algorithmic wage-setting, performance surveillance, and instant termination with no appeal or collective bargaining mechanism. 2. Communities subjected to AI surveillance in welfare distribution, law enforcement, and financial services, with no local auditing capacity or redressal mechanisms. 3. Traditional knowledge holders including artists, healers, farmers, whose cultural assets are scrapped to train commercial AI models without consent or compensation. 4. Civil society organizations from the region who are excluded from global norm-setting due to visa barriers, language exclusion, and lack of dedicated funding. 5. Non-English speaking users for whom AI systems fail on contextual accuracy, amplifying misinformation, and eroding digital trust. The Global Dialogue should prioritise: i. dedicated participation seats for worker representatives and affected communities, not just civil society intermediaries, with accessible participation infrastructure, such as translation support and direct funding. ii. community impact assessments before AI systems are deployed in public services or conflict zones. a benefit-sharing framework for data and knowledge extracted from South Asia

1. Mandated documentation: Every session must produce a 1-page summary stating explicitly where convergence exists and where it does not. This prevents ambiguous outcomes and gives non-state actors a record to advance their positions in subsequent processes. 2. Problem-solving workshops: Mixed stakeholder groups, data workers, regulators, tech firms, civil society, receive a real governance dilemma (e.g., "How to provide appeal rights for algorithmic terminations") and 90 minutes to draft a recommendation. This breaks silos and forces trade-off conversations. 3. Community-led listening sessions: Affected communities present evidence of harm directly to policymakers with no intermediary reframing. Session formats must reserve time for policymakers to respond publicly to what they heard, shifting power toward lived experience. 4. Regional consultations: Existing mechanisms like the Asia Pacific Assembly, RightsCon, and IGF regional meetings are valuable for pre-Dialogue consultation. The dialogue should adopt the formats above to ensure these inputs translate into outcomes.

Effective AI governance requires concrete practices that make participation meaningful, not just symbolic. The Dialogue should adopt a dedicated online platform where background materials are published in advance, written feedback is transparently collected, and outputs include synthesis reports explaining how contributions shaped outcomes. Regional and intersessional consultations are equally critical. The Dialogue should engage with existing regional convenings so that conversations are grounded in locally specific experiences. Where internet access and visa barriers are acute, financial support, low-bandwidth options, multilingual materials, and safety measures are core governance enablers, not ancillary accommodations. Taken together, these practices shift participation from passive attendance to consequential engagement. The Dialogue's added value is to connect and adapt these existing models specifically to AI governance, reducing fragmentation while making global cooperation more inclusive and accountable.