University of the West of England, Bristol

Success cannot be measured by participation alone. The history of international governance is littered with well-attended dialogues that produced declarations without teeth. For this Dialogue to be genuinely successful, it must move from consultation to commitment, and from rhetoric to architecture. Five outcomes would mark a meaningful success: First, the adoption of a Structural Participation Framework that guarantees Global South representation not as a courtesy, but as a governance requirement in all AI standard-setting processes that follow from this Dialogue. Consultation without structural integration perpetuates the asymmetry the Dialogue is mandated to address. Second, agreement on the principle of algorithmic transparency rights: that individuals, educational institutions, and states have a right to know when AI systems are being used to assess, evaluate, or exclude them, and a right to contest those determinations. This must be named as a human rights imperative, not a technical recommendation. Third, a mandate for a binding multilateral instrument on AI governance in education — addressing admissions, assessment, and academic integrity with timelines, responsible bodies, and Global South co-authorship built in from the outset. Fourth, a Compute Equity Commitment: recognition that governance frameworks divorced from infrastructure inequality are cosmetic. Success requires commitments to redistributing AI computing capacity, not merely issuing principles about responsible use of systems Global South actors cannot access. Fifth, and most critically, the Dialogue must produce a follow-on mechanism with legal personality: not another working group, but a standing body empowered to monitor, report, and recommend enforcement. Words without institutions are not governance. The measure of success is simple: will the populations most harmed by unregulated AI be more protected after this Dialogue than before it? That is the standard to apply.

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

These four priorities reflect the structural arguments advanced in this submission and the urgent governance realities documented in comparative research across Africa, the Middle East, and Southeast Asia. Protection and promotion of human rights is the foundational priority. The "equality of arms" framework advanced here reframes AI governance not as a technical challenge but as a rights obligation. When AI systems deployed in educational settings systematically disadvantage learners from the Global South in admissions, assessment, and academic integrity monitoring, this is not a malfunction. It is a human rights failure requiring international remedy. Transparency, accountability, and human oversight is an urgent corollary. No binding international instrument currently creates a right for students or institutions to know when AI is being used to evaluate or exclude them, or to challenge those determinations. This accountability gap is particularly acute in low-resource environments where affected parties lack the legal or technical capacity to identify, let alone contest, algorithmic decisions. The Dialogue is uniquely positioned to establish minimum transparency standards with global reach. AI capacity-building cannot be separated from governance. With less than 0.5% of global AI computing capacity located on the African continent, governance frameworks that do not address infrastructure inequality risk encoding the very divides they purport to close. Meaningful participation in AI governance requires the capacity to develop, deploy, and contest AI systems, not merely to receive them. Social, economic, ethical, cultural, linguistic and technical implications of AI captures the multidimensional harm documented in this submission. The erosion of academic sovereignty, bias against non-English speakers, and inequitable credentialing outcomes are simultaneously social, ethical, cultural, and technical failures requiring governance responses that match their complexity. Together, these four priorities form a coherent, rights-centred, structurally inclusive governance agenda.

Three significant issues fall between or beyond the listed thematic areas and merit explicit attention in the Dialogue's architecture. First, the Participation Architecture Problem. The listed themes address what AI governance should cover but not who designs it. The structural exclusion of Global South actors from AI standard-setting processes is not captured by any single theme ,it cuts across all of them. A cross-cutting workstream on inclusive governance design, addressing representation, voting rights, and veto power in emerging AI governance bodies, is essential. Without it, even well-intentioned thematic discussions risk producing frameworks that replicate existing asymmetries. Second, AI and Academic Sovereignty. The intersection of AI deployment and institutional autonomy in education represents an emerging governance gap that sits uncomfortably across the human rights, social implications, and capacity-building themes without being fully addressed by any. When universities in developing countries become dependent on AI systems they did not build, cannot audit, and cannot afford to replace, a new form of technological dependency emerges, one with direct implications for epistemic sovereignty, research independence, and the long-term capacity of Global South institutions to participate in knowledge production on equal terms. This requires dedicated governance attention. Third, Temporal Justice in AI Governance. Current governance frameworks are calibrated to existing AI systems. Yet the populations most harmed today, those in low-resource, non-English-speaking, and developing-country contexts are precisely those least equipped to engage in governance processes with long legislative timelines. The Dialogue should consider mechanisms for interim protections: provisional standards, emergency transparency requirements, and rapid-response accountability tools that provide meaningful protection now, rather than after multi-year negotiation cycles conclude. These are not peripheral concerns. They are the conditions under which all other thematic progress will either succeed or fail.

This response speaks from two intersecting positions: as a researcher based in the United Kingdom, and as a scholar whose comparative work spans AI governance frameworks across Africa, the Middle East, and Southeast Asia, regions where the governance gaps identified above are not theoretical but immediately consequential. In the education sector globally, the absence of binding transparency and accountability standards is producing measurable harm. AI-powered admissions screening tools, automated essay scoring systems, and academic integrity platforms are being deployed at scale in universities across the Global South, often procured from vendors in the Global North, with no obligation to disclose training data, performance disparities, or appeal mechanisms. Institutions lack the legal frameworks to challenge these systems; students lack the rights to contest outcomes. Across Africa specifically, the compute inequality gap translates directly into research dependency. Universities cannot train locally relevant AI models, cannot audit systems affecting their students, and cannot participate meaningfully in the standard-setting processes that will govern AI for the next generation. The governance gap is simultaneously a capacity gap and a rights gap. In the Middle East and Southeast Asia, linguistic and cultural bias in AI systems deployed in education compounds existing inequalities. Assessment tools calibrated for Western academic English systematically underperform for multilingual learners — yet no international framework requires vendors to disclose or remediate this disparity. The opportunity, however, is significant. The Global Dialogue arrives at a moment when many Global South governments are actively developing national AI strategies and are hungry for internationally legitimate governance frameworks to anchor them. A Dialogue that produces actionable, binding minimum standards particularly on transparency rights and participatory governance would meet a genuine demand that existing frameworks from the OECD, EU, or G7 processes have structurally failed to address.

The AI Dialogue arrives at a rare inflection point. Existing international cooperation on AI governance has been fragmented, jurisdiction-specific, and structurally exclusionary. The OECD Principles, the EU AI Act, the G7 AI Hiroshima Process, and the Bletchley Declaration each represent meaningful advances, but none carries universal legitimacy, and none was designed with the Global South as a co-author rather than a recipient. The Dialogue's distinctive role lies in four areas where no existing mechanism is positioned to act. First, legitimacy production. Only a process anchored in the UN system with universal membership and UNGA mandate can produce governance norms that carry genuine international legitimacy across jurisdictions with divergent regulatory traditions. The Dialogue can create the normative foundation that bilateral and regional frameworks cannot. Second, interoperability architecture. Rather than competing with existing frameworks, the Dialogue can establish minimum floor standards on transparency, accountability, and participation rights that regional and national frameworks must meet, without mandating uniformity. This preserves regulatory diversity while closing the most dangerous governance gaps. Third, structural inclusion as a design principle. The Dialogue can model what genuine multilateral AI governance looks like, by building Global South co-authorship into its process architecture, not as an afterthought but as a design requirement. This sets a precedent for every governance process that follows. Fourth, a standing cooperation mechanism. Sustainable international cooperation requires institutions, not just instruments. The Dialogue should lay the groundwork for a permanent multilateral body, with monitoring, reporting, and advisory functions that maintains cooperation momentum beyond July 2026. The first Global Dialogue will be judged not by what it declares, but by what it builds. The infrastructure for durable, equitable international cooperation is the outcome that matters most.

The AI Dialogue does not enter a vacant field. Several existing initiatives provide foundational architecture it should connect with, while bringing added value that none of them individually can deliver. The OECD AI Principles and the EU AI Act represent the most operationally developed governance frameworks to date. The Dialogue should treat them as a floor, not a ceiling acknowledging their technical rigour while explicitly addressing their jurisdictional limitations and the exclusion of developing countries from their design processes. UNESCO's Recommendation on the Ethics of AI (2021) is the closest existing instrument to a universal normative framework and carries significant legitimacy across the Global South. The Dialogue should build directly on its ethical foundations while advancing from recommendation to binding obligation particularly on transparency, accountability, and educational equity. UNCITRAL's ongoing work on AI and emerging technologies offers a model of inclusive, technical multilateral law-making with developing country participation built into its mandate. The Dialogue should draw on UNCITRAL's methodological rigour and its experience translating complex technical realities into legally operative instruments. The ITU's AI for Good platform provides an established multi-stakeholder convening infrastructure and a direct link to the Global South technology and development community. Connecting the Dialogue's normative outputs to ITU's implementation networks would accelerate uptake in precisely the jurisdictions most in need of governance support. The UN Secretary-General's Roadmap for Digital Cooperation and the Global Digital Compact provide the broader institutional context within which the Dialogue sits. Coherence with these frameworks rather than duplication should be a deliberate design principle. The added value the Dialogue brings is irreplaceable: universal membership, UNGA mandate, and the political authority to convert existing recommendations into binding international commitments. That is the gap none of the above can fill alone.

Structure should serve substance. The first Dialogue sets the precedent for all that follow. Governments should arrive with mandates, not merely positions. The Dialogue's value is diminished if delegations lack authority to commit to follow-on mechanisms. Developed country governments in particular should come prepared to make concrete commitments on capacity-building, compute equity, and technology transfer, not simply to advocate for principles they already embody. Civil society and affected communities, particularly from the Global South must be given substantive speaking roles, not observation status. Structured civil society panels, guaranteed intervention rights, and written submission processes that are genuinely read and reflected in outputs would signal a departure from the performative inclusion that characterises too many international forums. Academia and independent researchers can provide the evidentiary foundation the Dialogue needs documenting harms, testing governance proposals against real-world conditions, and offering comparative analysis across jurisdictions. Dedicated academic input sessions, pre-circulated research briefs, and formal mechanisms to incorporate scholarly evidence into negotiated text would substantially improve output quality. The private sector brings technical knowledge that no governance process can afford to ignore, but must engage under conflict-of-interest safeguards, with transparency about commercial interests and mandatory disclosure of AI system performance data relevant to governance discussions. On format, the Dialogue should avoid plenary-heavy structures that reward prepared statements over genuine exchange. Smaller, thematically focused working groups with mixed stakeholder composition, co-facilitated by Global South and Global North representatives — will produce more substantive outputs than large ceremonial sessions.

Four categories of underrepresentation demand urgent structural remedy. First, Global South governments and institutions. Africa, Southeast Asia, the Pacific Islands, and large parts of Latin America and the Middle East remain peripheral to AI governance standard-setting. Inclusion requires more than invitation, it requires funded participation, advance access to negotiating texts, technical support for delegations, and co-facilitation rights in working groups. Presence without capacity is not representation. Second, non-English-speaking communities. The language of global AI governance is overwhelmingly English, a structural bias that filters out not only linguistic communities but entire epistemic traditions. Multilingual submission processes, interpretation support, and explicit recognition of non-Western governance philosophies in normative frameworks are minimum requirements for genuine inclusion. Third, affected learners and educational communities. Students subjected to AI-driven admissions screening, automated assessment, and algorithmic academic integrity monitoring have no formal standing in any existing AI governance forum. Youth advisory mechanisms, structured student input processes, and education sector representation on governance bodies would begin to close this gap. Fourth, indigenous and local communities. These communities face distinct AI governance challenges, including data extraction without consent, cultural misrepresentation in training datasets, and the erosion of knowledge sovereignty. Free, prior, and informed consent principles from environmental and cultural rights law should be explicitly applied to AI data governance affecting indigenous communities. Inclusion is not achieved by opening doors. It requires redesigning the room, ensuring that underrepresented voices shape agendas, co-author outputs, and hold accountability rights over processes that will govern technologies already reshaping their lives.

Traditional multilateral formats i.e. prepared statements, plenary debates, and closed drafting sessions are poorly suited to the complexity and urgency of AI governance. The first Global Dialogue has an opportunity to model a new kind of international engagement. Three innovative formats would make a measurable difference. Structured Adversarial Panels - Rather than sequential presentations, pair governance perspectives in direct dialogue, a Global South regulator with a Global North technology company; an affected student with a platform deploying AI in admissions. Structured disagreement, facilitated rigorously, produces sharper governance outputs than consensus-seeking panels that obscure real tensions. The goal is not conflict for its own sake, but intellectual honesty about where interests genuinely diverge. Living Document Drafting Sessions - Publish draft governance principles in real time during the Dialogue, open to tracked amendments from all registered stakeholders through a transparent, visible process. This transforms passive observers into active contributors and creates immediate accountability for whose input is accepted or rejected. It also models the kind of transparent, participatory governance the Dialogue itself is advocating for AI systems. Scenario Simulation Exercise- Present delegations with concrete, realistic AI governance failure scenarios, an automated admissions system denying thousands of qualified applicants in a developing country; a compute outage disproportionately affecting Global South researchers and require working groups to produce governance responses in real time. Scenario-based engagement surfaces operational gaps that abstract principle discussions routinely miss. Asynchronous Global Input Mechanisms - Not every meaningful voice can travel to Geneva. Pre-Dialogue digital submission portals, translated into at least six UN languages, combined with structured synthesis reports presented at the opening session, would ensure that the conversation in the room reflects the full breadth of global experience, not just those with travel budgets.

The global AI governance landscape, while fragmented, contains instructive examples of policies and practices that the Dialogue should study, adapt, and build upon, particularly where they demonstrate replicability across different regulatory traditions and resource contexts. The EU AI Act's risk-tiered regulatory architecture offers the most operationally developed model for binding AI oversight. Its classification of high-risk applications including AI used in education, employment, and essential services combined with conformity assessment requirements and transparency obligations, provides a legislative template that other jurisdictions can adapt. Its limitation is jurisdictional; its logic is transferable. UNESCO's Recommendation on the Ethics of AI demonstrates that normative consensus across 193 member states is achievable. Its ethical impact assessment methodology and emphasis on cultural diversity provide a foundation the Dialogue should elevate from recommendation to enforceable commitment. Rwanda's National AI Policy represents one of the most sophisticated Global South governance frameworks, explicitly integrating AI development with national equity goals and SDG alignment. It demonstrates that developing countries are not governance laggards, they are governance innovators whose approaches deserve structural recognition in international frameworks. The UNCITRAL model law approach ,developing flexible, jurisdiction-neutral legislative templates that states can adopt and adapt, offers a proven multilateral law-making methodology directly applicable to AI governance. Its track record in harmonising commercial law across divergent legal systems makes it a natural model for AI transparency and accountability standards. Singapore's Model AI Governance Framework demonstrates how voluntary, sector-specific guidance can achieve meaningful industry compliance without prescriptive regulation particularly valuable for jurisdictions building governance capacity incrementally. Across these examples, the common thread is specificity: effective AI governance names concrete obligations, assigns clear responsibilities, and creates meaningful accountability mechanisms.