Supreme Court of India

The first Global Dialogue on AI Governance will be a success if it produces three concrete outcomes. First, an agreed statement of foundational principles -- grounded in existing international human rights law -- that affirms the applicability of the UDHR, ICCPR, and ICESCR to AI systems without creating parallel or weaker substitute frameworks. Second, a clear and inclusive roadmap for the 2027 New York session, with structured mechanisms for the formal participation of civil society, academia, and technical experts from developing countries, not merely as observers but as co-contributors to substantive outputs. Third, a commitment by participating states to develop a common international lexicon for AI governance -- defining terms such as high-risk AI, general purpose AI, and algorithmic decision-making -- to prevent regulatory fragmentation and enable meaningful cross-border cooperation. From the perspective of LexMentor Legal Research and Policy Advisory, a research institution in India, success also requires that the Dialogue explicitly addresses the concerns of the Global South. The majority of the world's population lives in countries that are primarily consumers of AI systems developed elsewhere, with limited capacity to shape the design, training data, or governance of those systems. If the first session produces outcomes that reflect only the interests and regulatory philosophies of high-income countries and large technology companies, it will have failed its foundational mandate of open, transparent, and inclusive discussion. Finally, success requires that the Dialogue produce at least one actionable deliverable -- whether a set of common principles, a framework for mutual recognition of conformity assessments, or a proposal for a permanent intergovernmental panel on AI -- that gives the process tangible credibility and builds momentum toward the 2027 session.

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

LexMentor selects these four thematic areas as priorities because they most directly reflect the challenges and opportunities facing India and the broader Global South. Protection and promotion of human rights is the non-negotiable foundation of all legitimate AI governance. AI systems are already making or significantly influencing decisions in criminal justice, social welfare, credit, healthcare, and employment -- domains where errors or biases can cause irreversible harm to individuals. International governance must establish binding accountability standards and effective remedy mechanisms, not aspirational principles alone. Transparency, accountability, and human oversight are urgent priorities because the diffusion of responsibility across complex AI value chains -- developers, deployers, distributors, and users -- is creating a serious accountability gap. Individuals harmed by AI systems frequently have no identifiable responsible party. Minimum standards for explainability, human-in-the-loop oversight for high-stakes decisions, and accessible redress mechanisms are essential. AI capacity-building is a priority because the governance gap between high-income and low- and middle-income countries is widening. India and other emerging economies are increasingly dependent on AI systems developed abroad for critical public services, yet lack the institutional capacity, regulatory expertise, and computational infrastructure to govern those systems effectively or develop competitive alternatives. Meaningful international cooperation must include technology transfer, open-source model commitments, and investment in research and development capacity in the Global South. Social, economic, ethical, cultural, and linguistic implications of AI are priorities because AI systems trained predominantly on high-income country data systematically underperform for populations in the Global South -- in local languages, cultural contexts, and socioeconomic conditions. The governance framework must address not only technical safety but the structural inequalities that AI deployment risks entrenching.

Three cross-cutting issues deserve urgent attention that are not fully captured by the listed themes. First, AI and the rule of law. The deployment of AI in legal and judicial systems -- including predictive policing, automated bail and sentencing recommendations, AI-assisted legal research, and algorithmic case management -- poses profound challenges for the right to a fair trial, the right to an effective remedy, and the independence of the judiciary. These issues sit at the intersection of human rights, transparency, and social implications, but require a dedicated governance track addressing minimum standards for AI in justice systems. Second, generative AI and information integrity. The rapid proliferation of AI-generated text, images, audio, and video is already affecting electoral processes, judicial proceedings, and public trust in institutions worldwide. Mandatory provenance labelling, deepfake governance, and platform transparency obligations are urgent governance needs that cut across all thematic areas and require specific international standards. Third, data sovereignty and the governance of training data. The question of whose data is used to train AI systems, under what conditions, and with what benefit-sharing arrangements for the communities that generated that data, is a fundamental governance issue that is not adequately addressed by any of the listed themes. This is particularly acute for indigenous communities, whose cultural heritage, languages, and knowledge systems are increasingly being incorporated into AI training datasets without consent or compensation. The AI Dialogue should establish a dedicated workstream on data governance, data sovereignty, and the rights of data-generating communities.

India presents a vivid illustration of the governance gaps in global AI frameworks. As a country of 1.4 billion people undergoing rapid digital transformation, India is simultaneously a major market for AI systems, a growing AI developer, and home to some of the world's most complex social, linguistic, and economic diversity. The most significant challenge is the absence of binding accountability standards for AI systems deployed in public administration. India's welfare delivery systems -- including food security, employment guarantee schemes, and housing benefits -- are increasingly mediated by algorithmic systems that make or influence eligibility decisions affecting hundreds of millions of people. When these systems produce errors or discriminatory outcomes, affected individuals typically have no effective remedy. Existing domestic legal frameworks were not designed for this challenge and international governance has not filled the gap. In the legal sector specifically, where LexMentor operates, AI tools are transforming legal research, document review, and access to justice. This creates significant opportunities -- particularly for expanding access to legal assistance for underserved populations -- but also serious risks, including the perpetuation of judicial biases encoded in training data drawn from historical case law. The linguistic diversity of India -- with 22 scheduled languages and hundreds of dialects -- means that AI systems trained predominantly on English-language data consistently underperform for the majority of the population. This is not merely a technical inconvenience; it is a structural inequality that AI governance must address. The opportunity is equally significant. India's scale, its vibrant legal and policy research community, and its experience governing technology at population scale mean that it has valuable lessons to contribute to the international AI governance process -- if that process is genuinely inclusive.

The AI Dialogue can play three distinct and irreplaceable roles in advancing international cooperation. First, it can serve as the primary multilateral forum where developing countries have an equal voice in shaping international AI governance norms -- a role that no existing mechanism adequately fulfils. The OECD AI Principles, the G7 Hiroshima Process, and most AI safety summits to date have been driven by high-income countries. The AI Dialogue, operating under the auspices of the UN General Assembly with universal membership, has the unique legitimacy to produce governance frameworks that reflect the full diversity of the international community. Second, the Dialogue can serve as a clearinghouse and harmonisation mechanism for the proliferating landscape of national and regional AI governance frameworks. The EU AI Act, India's forthcoming Digital India Act, Brazil's AI Bill, the African Union's Continental AI Strategy, and dozens of other national frameworks are developing in parallel, creating significant fragmentation. The AI Dialogue can identify common minimum standards, facilitate mutual recognition, and reduce the compliance burden on developers and deployers operating across multiple jurisdictions. Third, the Dialogue can catalyse international investment in AI governance capacity, particularly for low- and middle-income countries. By establishing a formal capacity-building workstream and connecting it to existing UN development finance mechanisms, the Dialogue can help bridge the governance gap between high-income and developing countries -- ensuring that the latter are not merely subject to AI governance frameworks they had no role in designing.

The AI Dialogue should build upon and connect with several existing initiatives, while adding distinct value to each. The UNESCO Recommendation on the Ethics of Artificial Intelligence (2021) -- adopted by all 193 UNESCO Member States -- provides the most broadly endorsed normative foundation for international AI governance and should be explicitly recognised as the baseline ethical framework for the AI Dialogue's deliberations. The Global Partnership on AI (GPAI) has developed significant technical and policy expertise through its working groups on responsible AI, data governance, and the future of work. The AI Dialogue should incorporate GPAI's research outputs and invite its members to contribute to the technical dimensions of the Dialogue's agenda. The ITU AI for Good platform and the AI for Good Global Summit provide an established mechanism for connecting AI governance discussions with concrete development applications. The AI Dialogue should build a formal bridge to this platform to ensure that its outcomes are grounded in practical implementation experience. The Council of Europe's Framework Convention on AI (2024) -- the first binding international AI treaty -- offers a valuable precedent for translating governance principles into enforceable legal obligations. The AI Dialogue should study this instrument carefully, while recognising that a global instrument must be more attentive to the concerns of developing countries than the Council of Europe process was. The added value that only the AI Dialogue can bring is its universal membership, its location within the UN system, and its mandate to be genuinely inclusive of all stakeholders -- including civil society, academia, indigenous communities, and the private sector -- on equal terms with governments.

The AI Dialogue should adopt a multi-stakeholder model that goes beyond token consultation and gives non-governmental actors a substantive role in shaping outcomes. For the July 2026 Geneva session, LexMentor recommends a three-track structure. A government track for formal intergovernmental deliberation, with regional group coordination to ensure that developing country positions are effectively aggregated and represented. A stakeholder track for civil society, academia, the private sector, and technical experts, with designated speaking time in plenary and full participation in thematic working groups. A youth and innovation track specifically for early-career researchers, students, and young professionals from developing countries, recognising that they will live longest with the consequences of today's governance decisions. Written inputs, such as this submission, should be made publicly available on the UN website in all six official languages, and a summary synthesis of all inputs should be prepared by the Secretariat and distributed to all participants before the Geneva session. The working group format should be preferred over general debate for substantive deliberation. Small, thematically focused working groups with balanced geographic and stakeholder representation are more likely to produce actionable outputs than large plenary sessions. Simultaneous interpretation in all six official UN languages is essential for meaningful participation by delegations and stakeholders from non-English-speaking countries. Interpretation services should also be made available for regional languages where feasible, particularly for the African and South Asian regional groups.

Several communities are systematically underrepresented in global AI governance discussions and must be actively included in the AI Dialogue. Indigenous communities are perhaps the most underrepresented. Their languages, cultural heritage, and traditional knowledge are being incorporated into AI training datasets at scale, often without consent or compensation, yet they have virtually no voice in AI governance processes. The AI Dialogue should establish a dedicated indigenous peoples' consultation mechanism, consistent with the UN Declaration on the Rights of Indigenous Peoples, and should invite the UN Permanent Forum on Indigenous Issues to contribute to the Dialogue's deliberations. Civil society organisations and legal researchers from low- and middle-income countries are significantly underrepresented relative to their counterparts in high-income countries, primarily because of resource constraints that prevent participation in international processes. The AI Dialogue should establish a dedicated fellowship or travel grant programme to enable participation by researchers and civil society actors from the Global South. Persons with disabilities are major users of assistive AI technologies and are disproportionately affected by AI systems that fail to account for accessibility needs. The AI Dialogue should ensure that disability rights organisations are represented in all thematic working groups, and should require that all Dialogue outputs address the specific implications of AI governance for persons with disabilities. Women and girls, particularly in the Global South, face distinctive risks from AI systems -- including algorithmic bias in economic opportunity, AI-enabled gender-based violence through deepfake technology, and the underrepresentation of women's perspectives in AI training data. A gender-responsive approach should be mainstreamed across all thematic areas of the Dialogue.

LexMentor recommends three innovative engagement formats that could make the AI Dialogue genuinely dynamic and inclusive. First, a Global Citizens' Jury on AI Governance -- convened in parallel with the July 2026 Geneva session -- in which randomly selected members of the public from diverse countries deliberate on AI governance priorities and present their recommendations directly to the intergovernmental session. Citizen deliberation processes have been used effectively in Ireland, France, and several other countries to build public legitimacy for complex policy decisions, and an international version would signal that AI governance is genuinely a matter for all humanity, not just governments and corporations. Second, regional pre-consultations in the months preceding the Geneva session, organised in partnership with regional bodies such as the African Union, ASEAN, SAARC, and CELAC, to enable structured regional input into the Dialogue's agenda. These consultations should be conducted in regional languages and their outputs should be formally incorporated into the Geneva session's preparatory documentation. Third, a real-time annotation and response mechanism for the AI Dialogue's official documents -- modelled on the open consultation processes used by the Internet Governance Forum -- that allows registered stakeholders to submit comments and suggested amendments to draft texts during the Geneva session itself. This would create a transparent, iterative drafting process that is responsive to stakeholder feedback in real time, rather than presenting stakeholders with a finished text that they can only endorse or reject. All session recordings should be made publicly available with accurate captions and transcripts in multiple languages, and a post-session public report should be published summarising the key discussions, areas of agreement, and outstanding issues for the 2027 New York session.

Several policies and practices from India and internationally offer concrete lessons for effective AI governance. India's Aadhaar biometric identity system, while controversial, demonstrates both the transformative potential and the serious risks of large-scale AI-mediated public service delivery. The Supreme Court of India's decision in Justice K.S. Puttaswamy v. Union of India (2017) -- recognising the right to privacy as a fundamental right under the Constitution -- provides a powerful example of how constitutional courts can establish binding governance standards for AI and data systems in the absence of specific legislation. This judicial model of rights-based AI accountability deserves international attention. The European Union's AI Act (2024) is the most comprehensive binding AI governance framework currently in operation and offers valuable lessons in risk-based regulation, mandatory requirements for high-risk AI systems, and the institutional architecture needed to enforce AI governance at scale. However, its focus on systems developed and deployed within the EU means that its governance of cross-border AI impacts is limited -- a gap that the AI Dialogue should address. The Algorithmic Accountability Act proposals in the United States Congress, while not yet enacted, have generated a significant body of technical and policy analysis on algorithmic impact assessments that the AI Dialogue should draw upon. At the multilateral level, UNESCO's Recommendation on the Ethics of AI and its associated Readiness Assessment Methodology -- which has been applied in over 40 countries -- provide a practical tool for building AI governance capacity and assessing gaps that the AI Dialogue should formally adopt and expand. Finally, India's Digital Public Infrastructure approach -- building open, interoperable, privacy-preserving digital systems as public goods -- offers a model for AI infrastructure governance that prioritises public benefit over private capture and is directly replicable in other developing countries.