FINTECH

First, a shared definitional foundation. Governments currently approach AI governance with divergent understandings of core concepts "trustworthy AI," "human oversight," "high-risk systems." The first session should produce an agreed working lexicon that developing nations helped shape, not one imported wholesale from frameworks designed in Brussels or Washington. Without this, interoperability of governance approaches remains aspirational. Second, a concrete capacity commitment with accountability. Bridging AI divides requires more than acknowledgement. A successful Dialogue would produce specific, time-bound commitments from technologically advanced nations and multilateral institutions on compute access, open model availability, and regulatory technical assistance with a named follow-up mechanism. Statements of intent that carry no reporting obligation have characterised too many previous digital governance forums. Third, genuine integration of practitioner voices from the Global South. Policymakers in Geneva and New York rarely hear from the loan officer in Lahore, Pakistan assessing an SME's creditworthiness, or the trade finance practitioner navigating correspondent banking de-risking in Karachi. These are the frontlines where AI governance decisions land in practice. A successful Dialogue would establish standing channels not one-off consultations for practitioners, civil society, and private sector actors from developing countries to feed operational reality into ongoing governance work. The first session cannot resolve every tension. But it can establish that this forum is structurally different from its predecessors: universal in participation, grounded in practice, and accountable to outcomes. That foundation would itself be a success worth building on.

• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches
• Transparency, accountability, and human oversight

These four priorities reflect the operational reality of AI adoption in emerging market financial systems, where governance gaps are not theoretical they are already shaping who benefits and who is excluded. AI capacity-building is foundational. Across South Asia and sub-Saharan Africa, financial institutions are adopting AI-driven credit scoring, fraud detection, and compliance tools built on datasets and risk models calibrated for advanced economies. Without deliberate capacity investment in technical skills, regulatory expertise, and compute access developing countries will remain consumers of governance frameworks rather than co-architects of them. Social, economic and cultural implications matter because AI in financial services does not affect all communities equally. SME borrowers, women entrepreneurs, and informal sector participants face compounded risks when algorithmic systems replicate historical exclusions. These implications require urgent, contextualised analysis not generalised frameworks applied uniformly across vastly different economic contexts. Interoperability of governance approaches is critical for trade finance specifically. Cross-border transactions pass through multiple regulatory jurisdictions. Fragmented AI governance regimes create compliance friction that disproportionately burdens smaller institutions and developing-country counterparties, ultimately restricting the trade flows these economies depend on. Transparency, accountability and human oversight underpin trust. In correspondent banking and SME lending, AI-assisted decisions affect livelihoods. Without explainability requirements and meaningful human review mechanisms, accountability disappears precisely where it is most needed. Effective AI governance must be grounded in these realities not designed around them after the fact.

AI and the future of correspondent banking and financial access. The listed themes do not adequately address AI's role in accelerating financial exclusion through automated compliance systems. Correspondent banking de-risking already withdrawing from developing-country markets due to perceived compliance risk is increasingly driven by AI-powered transaction monitoring and risk-scoring tools calibrated on advanced-economy data. The result is that entire national banking systems face restricted access to international trade finance infrastructure. This is an AI governance issue with immediate, measurable consequences for development financing, yet it sits between existing thematic clusters without being owned by any. Algorithmic sovereignty and data asymmetry. Developing nations are generating vast quantities of financial, agricultural, and behavioural data that trains AI systems they neither own nor govern. The Dialogue's current framing addresses capacity-building and open models but does not confront the structural question of who holds the data advantage and how that compounds over time. Algorithmic sovereignty the right of nations to meaningful participation in the AI systems that govern their citizens and economies deserves explicit treatment as a governance principle. AI governance and trade law coherence. As AI increasingly intermediates cross-border trade through document verification, compliance screening, customs classification, and financing decisions governance frameworks developed in AI-specific forums risk creating friction with existing WTO commitments and bilateral trade agreements. The Dialogue should actively engage with this interface rather than leaving incoherence to accumulate. A dedicated workstream connecting AI governance to trade law would prevent regulatory fragmentation from becoming an invisible barrier to developing-country trade participation.

How governance gaps are affecting Pakistan's financial sector and regional trade: Pakistan sits at a consequential intersection: a country with 240 million people, a large informal SME economy, and deep integration into global supply chains yet one largely absent from the rooms where AI governance decisions are made. The most immediate challenge is automated exclusion. International correspondent banks are deploying AI-driven compliance and transaction monitoring systems to manage de-risking decisions. These systems are trained on datasets that treat Pakistani financial institutions as inherently high-risk, regardless of individual institutional performance. The practical consequence is that Pakistani SMEs particularly exporters in textiles, surgical instruments, and agricultural commodities face restricted access to letters of credit, trade finance lines, and international payment infrastructure. AI is not creating this problem, but it is accelerating and entrenching it at scale. The second challenge is regulatory asymmetry. Pakistan's financial regulators are developing AI governance frameworks under the State Bank's fintech and digital banking licensing regime, but without access to the technical standards, red-teaming methodologies, or interoperability frameworks being developed in the EU, UK, and United States. Pakistani institutions must comply with AI-influenced correspondent bank requirements they had no role in shaping. The opportunity is significant but time-sensitive. Pakistan has a nascent but growing blockchain and fintech ecosystem, a young technically-skilled workforce, and a strategic position in regional trade corridors connecting Central Asia, China, and the Gulf. AI-enabled trade finance infrastructure if governed inclusively could dramatically reduce the documentary friction and compliance costs that currently disadvantage Pakistani exporters. The window to shape these outcomes through governance is now. Frameworks being established in 2026 and 2027 will determine whether Pakistan's financial sector is a participant in AI-enabled trade or a casualty of it.

The role of the AI Dialogue in advancing international cooperation on AI governance: The Dialogue's most valuable contribution will not be producing another framework. It will be becoming the legitimate forum where existing frameworks are made coherent, inclusive, and accountable to the countries most affected by their gaps. A translation layer between divergent regimes. The EU AI Act, NIST Risk Management Framework, and emerging national frameworks from India, Singapore, and others represent real progress but were designed in isolation. Developing-country institutions navigating cross-border transactions must satisfy multiple, sometimes contradictory compliance expectations simultaneously. The Dialogue is uniquely positioned to broker interoperability agreements that reduce this burden without flattening legitimate regulatory diversity. A permanent feedback loop from practice. International AI governance operates largely without systematic input from practitioners in developing economies the trade finance specialists, SME advisors, and regulators who observe governance consequences daily. Periodic consultations that feed into documents few practitioners see again are insufficient. The Dialogue should establish standing structured channels for practitioner input to inform ongoing work. Accountability for capacity commitments. Multilateral forums have repeatedly produced capacity-building pledges that dissipate without follow-through. A light but real reporting mechanism tracking specific commitments made by member states and institutions would distinguish this Dialogue from its predecessors. Establishing a durable principle. Most fundamentally, the Dialogue can legitimise that AI governance is a shared sovereign interest, not a technical matter resolved by the most capable actors and exported downward. Establishing that principle in a UN-mandated forum, with genuine developing-country co-authorship, would represent a structural shift worth more than any single agreement reached in Geneva or New York.

Existing initiatives the AI Dialogue should build upon: Several substantive efforts already exist. The Dialogue's added value lies not in duplicating them but in connecting them into a coherent, universally legitimate architecture. OECD AI Policy Observatory has built the most comprehensive comparative database of national AI governance approaches. The Dialogue should formally draw on this resource while ensuring that the observatory's coverage currently weighted toward OECD members is systematically extended to developing-country frameworks. FATF and correspondent banking compliance infrastructure represents an underappreciated entry point. AI-driven transaction monitoring systems deployed by international banks are already functioning as de facto governance instruments, determining which institutions and countries access global financial infrastructure. The Dialogue should engage FATF directly to ensure AI governance principles are embedded in the next revision of its recommendations. ITU's AI for Good platform and the UN Secretary-General's Roadmap for Digital Cooperation provide existing multilateral scaffolding for technical capacity-building. The Dialogue should consolidate rather than fragment these efforts, designating ITU as the operational delivery partner for compute access and technical assistance commitments made at the Dialogue level. The Global Partnership on AI (GPAI) produced valuable multistakeholder research before its transition into the OECD structure. Its working group methodology bringing together government, academia, civil society, and industry around specific problem statements is a model the Dialogue should adopt for its inter-sessional work. The added value the Dialogue uniquely brings is universal membership, UN legitimacy, and a mandate that explicitly includes governments currently absent from every initiative listed above. Its role is to be the forum where these efforts are accountable to all nations not just the ones that designed them.

Recommendations for stakeholder contribution and Dialogue structure: The Dialogue's legitimacy will ultimately rest on whether its outputs reflect the priorities of all nations not just those with the largest delegations and most resourced policy teams. Structural design must compensate for participation asymmetries that are otherwise self-reinforcing. Tiered input mechanisms. A single submission format systematically advantages well-resourced institutional actors. The Dialogue should offer differentiated pathways: formal written submissions for governments and multilateral bodies; shorter structured testimony formats for practitioners, SMEs, and civil society; and anonymised case submission channels for operational actors unable to engage publicly due to regulatory or institutional constraints. Standing multistakeholder advisory track. Episodic consultation between sessions is insufficient. A permanent advisory track with rotating membership explicitly weighted toward developing-country private sector, civil society, and technical community representatives should have formal agenda-setting rights, not merely observer status. The distinction matters: presence without influence reproduces the exclusion the Dialogue exists to correct. Regional preparatory convening. Before each plenary session, regionally organised preparatory meetings co-hosted with the African Union, ASEAN, OIC, SAARC, and comparable bodies would allow developing-country stakeholders to arrive with consolidated positions. Countries without permanent Geneva or New York missions are structurally disadvantaged in real-time negotiation; regional preparation partially offsets this. Mandatory input-to-output mapping. After each session, the Dialogue secretariat should publish a transparent account of how stakeholder submissions and testimony influenced agenda items, draft language, and commitments reached. Without this, participation feels performative and repeat engagement declines accordingly. The Dialogue's format should be designed around the question: who is currently absent from AI governance conversations, and what specific barrier does each structural choice remove

Underrepresented voices in AI governance and pathways to inclusion: Global AI governance discussions are currently shaped disproportionately by a narrow band of actors: large technology companies, regulators from OECD member states, and academics affiliated with institutions in North America and Europe. The perspectives most consequentially absent are precisely those of communities where AI governance decisions land hardest. Frontline financial sector practitioners in developing economies. Loan officers, trade finance specialists, and SME banking professionals in South Asia, sub-Saharan Africa, and the Middle East observe daily how AI-driven compliance systems, credit scoring models, and correspondent banking algorithms affect real livelihoods. Their operational knowledge is irreplaceable and almost entirely absent from governance forums. Structured practitioner testimony mechanisms, coordinated through national banking associations and regional development banks, would begin to close this gap. Informal economy participants. The majority of economic activity in developing countries occurs outside formal institutional frameworks. AI systems increasingly affect informal traders, smallholder farmers, and micro-entrepreneurs through credit access decisions, mobile money algorithms, and agricultural pricing platforms yet these communities have no meaningful representation in governance processes. Intermediary organisations with established community trust must be resourced and mandated to represent these voices. Linguistic and cultural communities outside the dominant governance languages. AI governance documentation, consultation processes, and technical standards are produced overwhelmingly in English. This structurally excludes vast populations across Arabic, Urdu, Swahili, Bengali, and hundreds of other language communities from meaningful participation. The Dialogue should mandate multilingual consultation materials and fund translation infrastructure as a core operating commitment, not an optional accessibility feature. Small and medium enterprise ecosystems globally. SMEs constitute the majority of employment and trade activity worldwide yet are almost invisible in AI governance deliberations dominated by large technology and financial sector incumbents. Inclusion requires resourcing, not just invitation.

Innovative engagement formats for meaningful AI Dialogue participation: Traditional plenary formats prepared statements delivered to half-empty rooms produce documentation, not dialogue. The AI Dialogue should deliberately adopt engagement formats that generate genuine exchange, surface operational knowledge, and produce outputs that inform rather than merely record. Practitioner testimony panels. Structured sessions where frontline practitioners trade finance officers, SME lenders, agricultural extension workers, customs officials from developing economies present specific cases of AI governance consequences they have directly observed. Not position papers. Not institutional statements. Operational testimony, cross-examined by technical experts and policymakers in real time. This format surfaces knowledge that no written submission process captures. Red team working sessions. Small, mixed groups combining government regulators, civil society critics, technical experts, and private sector practitioners tasked with identifying the specific failure modes of draft governance proposals before adoption. Borrowed from cybersecurity practice, this format stress-tests ideas rather than endorsing them, and produces more durable outputs as a result. Asynchronous regional deliberation tracks. Between sessions, moderated online deliberation platforms available in multiple languages and accessible via low-bandwidth connections would allow stakeholders who cannot travel to Geneva or New York to engage substantively with draft proposals over weeks rather than hours. Outputs from these tracks should carry formal weight in session agendas. Binding scenario exercises. Structured simulations where delegations navigate specific AI governance dilemmas a correspondent bank withdrawing from a developing-country market due to AI risk scoring; a cross-border trade finance dispute involving algorithmic document verification build shared understanding of governance consequences more effectively than abstract principle negotiation. Public accountability sessions. Each Dialogue session should close with a structured public review of commitments made at the previous session and progress achieved. Naming gaps explicitly, on the record, changes institutional behaviour.

Good practices and policy approaches in AI governance: Several initiatives offer concrete models worth scaling, adapting, and connecting through the Dialogue. Singapore's Model AI Governance Framework. Among the most practically useful national frameworks produced to date, Singapore's approach is notable for its sector-specific guidance, its focus on explainability requirements proportionate to risk level, and its deliberate design for adoption by organisations without large compliance teams. Its accessibility makes it a genuine model for developing-country regulators not merely an aspirational benchmark. The Bank for International Settlements' Project Nexus and related fintech frameworks. BIS innovation hub work on cross-border payment interoperability demonstrates that technical standards can be developed multilaterally with genuine developing-country participation. The methodology convening central banks from diverse economies around specific interoperability problems is directly transferable to AI governance standard-setting. Rwanda's national AI policy and Kenya's Blockchain and AI Taskforce. These represent underappreciated examples of developing-country governments proactively shaping AI governance rather than receiving it. Both demonstrate that resource constraints do not preclude thoughtful policy design and both deserve amplification and resourcing through the Dialogue rather than remaining isolated national experiments. The State Bank of Pakistan's regulatory sandbox framework. Pakistan's sandbox approach to fintech innovation allowing controlled deployment of new financial technology under regulatory supervision offers a practical model for governing AI applications in financial services under conditions of regulatory uncertainty. Sandbox methodologies allow governance to develop iteratively alongside technology rather than perpetually lagging behind it. GBBC's Digital Discretion framework and blockchain governance standards. Industry-led governance initiatives that establish transparency and accountability standards for blockchain-based financial infrastructure demonstrate that private sector actors can contribute binding governance commitments, not merely voluntary principles. The common thread across these examples is proportionality, inclusivity of design, and accountability to measurable outcomes.