Emerging Payments Association Asia

A successful first Global Dialogue on AI Governance would be defined by progress toward practical alignment that can be carried into implementation across jurisdictions. From an EPAA perspective, three outcomes are critical. First, interoperability established as an operational priority. This includes early convergence on shared terminology, and alignment in approaches to auditability, accountability, data, and compliance. Without this, AI-enabled systems - particularly in cross-border payments - risk fragmentation, higher costs, and reduced scalability across markets. Second, clearer foundations for trust in AI-enabled environments. The Dialogue should move beyond high-level principles to address how AI-driven decisions can be understood, tested, and relied upon across institutional and jurisdictional boundaries. Consistency in how AI is applied in areas such as fraud detection, identity, and compliance would strengthen confidence in deployment. Third, inclusion treated as a design requirement. Governance discussions should recognise that infrastructure and regulatory choices determine who can participate. Ensuring that frameworks remain accessible to smaller institutions and emerging markets is essential to avoid concentration of capability and to support broad-based adoption. Finally, the Dialogue should establish a basis for ongoing engagement between policymakers and industry practitioners. AI in payments is evolving through implementation, and governance will need to adapt alongside real-world deployment. Success would therefore be measured by whether the Dialogue enables alignment that systems can build on - supporting trust, interoperability, and inclusive participation in practice.

• Safe, secure and trustworthy AI
• Interoperability of governance approaches
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI

The selected priorities reflect an operational focus on how AI can be deployed in a way that supports trusted, interoperable, and inclusive systems, particularly in cross-border contexts. Safe, secure and trustworthy AI is foundational for payments and digital trade. Confidence in AI-enabled systems depends on their ability to operate reliably across jurisdictions, with consistent outcomes that users and institutions can depend on. Interoperability of governance approaches is critical to avoid fragmentation. Divergent regulatory and technical frameworks increase complexity and cost, particularly for cross-border transactions. Alignment in standards, terminology, and compliance approaches enables systems to function cohesively across markets. Transparency, accountability, and human oversight are necessary to support trust in practice. AI-driven decisions in areas such as fraud detection, risk management, and compliance must be explainable and auditable across institutional boundaries. This ensures that outcomes can be interpreted consistently and remain aligned with regulatory expectations. Social, economic, ethical, cultural, linguistic and technical implications of AI are directly linked to inclusion. Governance and infrastructure choices shape who can participate and under what conditions. Without consideration of these factors, there is a risk that AI adoption reinforces existing barriers, particularly for smaller institutions and emerging markets. Taken together, these priorities focus on enabling AI governance that is practical to implement across jurisdictions, supports trust in real-world deployment, and maintains broad participation in AI-enabled systems.

One cross-cutting issue that is not explicitly captured is the impact of agentic AI on transaction initiation and system design. Payment infrastructure has historically been built on the assumption that a human initiates each transaction, either directly or through an authorised agent. As AI systems increasingly act autonomously - initiating, routing, and optimising transactions - this assumption is being challenged. This raises practical governance questions around identity, authority, and liability: how AI agents are identified within payment systems, how their actions are authorised, and how responsibility is assigned when outcomes need to be reviewed or contested. A second emerging issue is the need for shared definitions and a common operational language across jurisdictions. While interoperability is referenced as a theme, the absence of aligned terminology and conceptual frameworks creates friction in implementation. Differences in how key concepts - such as "AI system," "risk," or "explainability" - are defined can lead to inconsistencies in regulatory interpretation and technical design, even where high-level objectives are similar. A third cross-cutting concern is the practical cost and complexity of compliance, particularly for cross-border participants. Governance approaches that do not account for operational realities risk creating disproportionate burdens for smaller institutions and those operating across multiple markets. This can limit participation and reduce the overall effectiveness of AI adoption. Addressing these issues would support governance frameworks that are not only principled, but also implementable across diverse systems and jurisdictions.

Governance gaps across the selected thematic areas are already shaping both the challenges and opportunities in AI-enabled payments, particularly in cross-border contexts. A primary challenge is fragmentation across jurisdictions. Divergent approaches to AI governance, including differences in regulatory expectations, definitions, and compliance requirements, increase operational complexity and cost. For payment providers operating across multiple markets, this creates inefficiencies in areas such as fraud management, risk scoring, and transaction monitoring, where AI models must be adapted to inconsistent standards. A second challenge relates to trust and interpretability. As AI becomes more embedded in decision-making processes, there is increasing pressure to ensure that outputs can be understood and validated across institutional and regulatory boundaries. Inconsistent expectations around transparency and auditability can limit the ability to scale AI solutions across markets. A third challenge is the risk of exclusion. Without alignment and proportionality in governance frameworks, compliance and technical requirements may become disproportionately burdensome for smaller institutions and participants in emerging markets. This risks reinforcing existing barriers to participation and limiting the reach of AI-enabled financial services. At the same time, these gaps present opportunities. There is a clear opportunity to drive interoperability by design, through greater alignment in standards, terminology, and governance approaches. This would support more efficient cross-border systems and reduce the need for post-implementation reconciliation. There is also an opportunity to embed trust as a system feature, by developing governance approaches that enable consistent, auditable, and reliable AI-driven outcomes across jurisdictions. Finally, addressing these gaps can support more inclusive participation, ensuring that the benefits of AI adoption extend beyond large institutions to a broader range of market participants.

The AI Dialogue can play a practical role in advancing international cooperation by supporting alignment that can be implemented across jurisdictions, rather than remaining at the level of high-level principles. First, it can act as a platform for developing shared understanding and common language. Differences in how key concepts are defined and applied create friction in cross-border implementation. Establishing greater consistency in terminology and baseline expectations would support more coherent governance approaches across markets. Second, the Dialogue can facilitate progress toward interoperability of governance frameworks. By identifying areas where regulatory and technical approaches can be aligned - such as auditability, accountability, and risk management - it can help reduce fragmentation and enable AI-enabled systems to operate more effectively across jurisdictions. Third, it can support structured engagement between policymakers and industry practitioners. AI governance challenges are increasingly emerging through real-world deployment. Ongoing dialogue with industry participants can help ensure that governance approaches reflect operational realities, including cost, scalability, and system constraints. Fourth, the Dialogue can help embed trust and inclusion into implementation pathways. This includes ensuring that governance approaches support transparent and reliable AI-driven outcomes, while remaining accessible to a broad range of participants, including smaller institutions and emerging markets. Finally, the Dialogue can establish a basis for continued international coordination, moving beyond one-off discussions toward iterative engagement. This is particularly important as AI technologies and their use in sectors such as payments continue to evolve. Through these roles, the AI Dialogue can support cooperation that translates into systems that are interoperable, trusted, and widely adoptable in practice.

The AI Dialogue should build on existing initiatives that are already addressing cross-border coordination, standards development, and practical implementation in financial services and digital infrastructure. One important foundation is the G20 Roadmap for Enhancing Cross-Border Payments, which focuses on improving speed, cost, transparency, and access. This work has already highlighted the importance of interoperability, regulatory alignment, and data standards - issues that are directly relevant to AI governance in payments. The Dialogue can also connect with international standard-setting bodies and industry-led frameworks that are developing approaches to data governance, risk management, and technical standards. These efforts provide a basis for more detailed alignment in how AI systems are designed, assessed, and deployed across jurisdictions. In addition, regional and industry forums - including those that convene financial institutions, fintechs, and infrastructure providers - offer insights from operational deployment. These mechanisms are valuable for understanding how governance approaches function in practice, particularly across diverse markets in Asia-Pacific. The added value of the AI Dialogue lies in its ability to bring these strands together at a global level. First, it can provide a coordinating layer, connecting policy discussions with technical and industry-led initiatives, reducing duplication and supporting coherence across frameworks. Second, it can accelerate alignment across jurisdictions, particularly where existing initiatives are regionally or sectorally focused. Third, it can elevate practical implementation considerations, ensuring that governance approaches are informed by real-world constraints such as cost, scalability, and cross-border complexity. Finally, it can help link global principles to operational pathways, supporting outcomes that are not only aligned in intent but workable across different systems and markets.

Different stakeholders can contribute most effectively to the AI Dialogue where their roles are clearly structured and connected to implementation. Governments and regulators can provide direction on policy objectives, legal frameworks, and areas where alignment is feasible. Their role is critical in identifying where convergence can reduce fragmentation across jurisdictions. Industry participants - including financial institutions, fintechs, and technology providers - can contribute insights from real-world deployment. This includes how AI systems operate in practice, where governance creates friction, and what is required for scalability across markets. Standard-setting bodies and technical organisations can support the development of common frameworks, terminology, and approaches to auditability, interoperability, and risk management. Civil society and academia can contribute perspectives on inclusion, societal impact, and long-term implications, ensuring that governance approaches remain balanced and broadly applicable. In terms of format and structure, the Dialogue would benefit from a multi-layered approach. First, thematic working tracks aligned to priority areas (such as interoperability, trust, and inclusion) can enable focused and technical discussion. Second, cross-sector working groups should bring together policymakers and practitioners to test how proposed approaches function in operational settings, particularly for cross-border use cases. Third, the Dialogue should include iterative engagement cycles, moving from discussion to draft outputs, followed by consultation and refinement. This supports continuity and avoids one-off outcomes. Fourth, regional input channels are important to reflect diverse market conditions and avoid a concentration of perspectives from a limited set of economies. Finally, outputs should be practically oriented, such as guidance, reference frameworks, or common definitions, which stakeholders can adopt and build upon across jurisdictions.

Several perspectives remain underrepresented in global AI governance discussions, particularly those connected to implementation in diverse and resource-constrained environments. Participants from emerging markets and smaller economies are often underrepresented, despite being directly affected by cross-border governance decisions. Payment systems and digital infrastructure in these markets operate under different constraints, and without their input, frameworks risk reflecting the assumptions of a limited set of advanced economies. Small and medium-sized institutions and fintechs are also less visible in global discussions. Governance approaches that are workable for large, well-resourced organisations may create disproportionate compliance and technical burdens for smaller participants, limiting their ability to adopt and scale AI solutions. Operational and technical practitioners - those responsible for implementing AI systems in areas such as payments, fraud, and compliance - are often not systematically included. Their insights are critical to understanding how governance functions in practice, including cost, interoperability, and system constraints. Cross-border industry bodies and ecosystem convenors can also play a stronger role. These organisations bring together diverse stakeholders and can surface practical challenges that may not be visible within single jurisdictions or sectors. To address these gaps, the AI Dialogue could: * Establish structured regional input channels, ensuring consistent participation from Asia-Pacific, Africa, and Latin America * Include targeted participation mechanisms for SMEs and fintechs, such as dedicated working groups or consultation tracks * Integrate practitioner-led sessions, focused on real-world implementation challenges * Leverage industry associations and convening bodies to aggregate and represent a broader range of voices Broadening participation in these ways would support governance approaches that are more representative, practical, and implementable across diverse systems and markets.

Meaningful engagement in the AI Dialogue will depend on formats that move beyond formal statements toward structured, interactive exchange grounded in implementation. One effective approach is scenario-based working sessions, where participants from different stakeholder groups respond to common cross-border use cases. This allows policymakers, industry, and technical experts to test how governance approaches operate in practice, surface points of friction, and identify where alignment is feasible. A second format is moderated, multi-stakeholder roundtables with defined outputs. Smaller, curated groups can enable more detailed discussion across jurisdictions and sectors, particularly when focused on specific issues such as interoperability, auditability, or inclusion. These sessions are most effective when designed to produce tangible outputs, such as draft principles or areas of convergence. A third approach is live polling and structured feedback mechanisms embedded within sessions. This enables a broader set of participants to contribute perspectives in real time, including those who may not be speaking directly. Aggregated responses can help identify areas of consensus and divergence quickly. Practitioner-led deep dives can also play an important role. Sessions led by those implementing AI systems can provide insight into operational realities, including cost, scalability, and cross-border complexity, grounding discussions in real-world deployment. Finally, iterative engagement cycles - where initial discussions are followed by refinement and re-engagement - can support continuity and progress over time, rather than one-off exchanges. These types of formats have proven effective in convening diverse stakeholders across regions and sectors, enabling practical alignment and ensuring that discussions translate into implementable outcomes.

A number of existing approaches across policy, industry, and technical domains offer practical pathways for effective AI governance. One example is the use of risk-based regulatory frameworks, where obligations are calibrated according to the level of risk posed by an AI application. This supports proportionality and enables innovation while maintaining safeguards, particularly relevant in areas such as fraud detection and compliance in payments. Common data and messaging standards in financial services provide another practical foundation. These approaches enable interoperability across institutions and jurisdictions by aligning how information is structured and exchanged. Extending similar principles to AI governance - particularly around data, identity, and auditability - can support more consistent cross-border implementation. Model governance and audit frameworks used within financial institutions also offer concrete practices. These include documentation of model behaviour, validation processes, and ongoing monitoring. Such approaches support transparency and accountability, and can be adapted to meet regulatory expectations across different markets. Industry-led collaboration platforms provide a further example. By convening stakeholders across sectors and regions, these platforms enable the sharing of operational insights and help align approaches to common challenges, particularly in cross-border contexts. Finally, public-private engagement mechanisms - where policymakers and industry participants work together on emerging issues - have proven effective in bridging the gap between high-level principles and implementation. These approaches allow governance frameworks to be informed by real-world constraints such as cost, scalability, and system design. Taken together, these examples highlight that effective AI governance is supported by approaches that are interoperable, proportionate, and grounded in operational practice, enabling adoption across diverse jurisdictions and market participants.