YAPEIM HOLDINGS SDN. BHD.

A successful first Global Dialogue on AI Governance should deliver outcomes that are both normative (shaping principles) and operational (enabling implementation). First, it should establish a shared baseline of governance principles that are internationally credible yet adaptable across jurisdictions. This includes alignment on core issues such as transparency, accountability, safety, and human oversight, while allowing room for cultural, legal, and developmental diversity. A meaningful success indicator would be early convergence, not uniformity. Second, the dialogue should produce a practical governance framework or reference model. Beyond high-level declarations, this could take the form of a modular architecture covering data governance, model governance, risk classification, and audit mechanisms. If policymakers and institutions can directly map this into their national strategies, the dialogue has achieved real utility. Third, it should catalyse multi-stakeholder coalitions. AI governance cannot be state-led alone. Success would involve concrete partnerships between governments, industry, academia, and civil society, with clear mandates such as pilot sandboxes, cross-border data initiatives, or shared testing standards. Fourth, the dialogue should address Global South inclusion and equity. This means ensuring that emerging economies are not merely participants but co-shapers of governance norms. Mechanisms such as capacity-building commitments, technology transfer pathways, and equitable data representation would be critical signals of success. Fifth, it should define a continuity mechanism. A one-off event has limited value. Establishing a standing working group, annual reporting structure, or implementation taskforce would ensure that dialogue translates into sustained progress. Ultimately, success lies in moving from conversation to coordinated action. If the dialogue results in shared principles, deployable frameworks, active partnerships, and a clear path forward, it would mark a strong foundation for responsible and globally inclusive AI governance.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Open-source software, open data and open AI models
• AI capacity-building

The selected priorities reflect a pragmatic and inclusive approach to AI governance, particularly from the perspective of emerging and developing ecosystems. Safe, secure and trustworthy AI is foundational. Without baseline safety, robustness, and risk management, trust in AI systems cannot be established, limiting both adoption and societal benefit. This priority ensures that AI deployment aligns with public interest and institutional accountability. AI capacity-building is critical to address global asymmetries. Many countries face structural gaps in talent, infrastructure, and institutional readiness. Strengthening local capabilities enables meaningful participation in the AI economy and prevents long-term dependency on external technologies. The broader social, economic, ethical, cultural, linguistic, and technical implications of AI must be actively addressed to ensure that AI systems are context-aware and culturally aligned. Governance frameworks should not be purely technical but must reflect societal values, diversity, and local realities, especially in multilingual and plural societies. Open-source software, open data, and open AI models play a key role in democratizing access to AI innovation. Openness reduces barriers to entry, accelerates research and development, and allows smaller actors, including public institutions and startups, to build and adapt AI solutions responsibly. Collectively, these priorities emphasize that effective AI governance is not only about control and regulation, but also about enabling equitable participation, fostering trust, and ensuring that AI delivers meaningful societal value across different contexts.

Yes. While the listed themes are comprehensive, several cross-cutting and emerging issues warrant more explicit attention. First, data sovereignty and equitable data governance remain under-articulated. Control over data flows, ownership, and value capture is increasingly central to national competitiveness and digital dignity, particularly for developing economies. Without fair data arrangements, AI risks reinforcing extractive dynamics where value accrues disproportionately to a few global actors. Second, compute and infrastructure inequality is an emerging structural constraint. Access to high-performance computing, cloud infrastructure, and energy resources is unevenly distributed, shaping who can meaningfully develop and deploy advanced AI. Governance discussions should therefore include mechanisms for shared infrastructure, federated access models, or regional compute alliances. Third, AI for public good and mission-oriented deployment deserves stronger emphasis. Beyond risk mitigation, there is a need to prioritise AI applications in areas such as social protection, climate resilience, healthcare, and financial inclusion. This shifts governance from a defensive posture to a developmental one. Fourth, value-aligned and culturally grounded AI is often implicit but insufficiently operationalised. Existing frameworks tend to be rooted in dominant epistemologies. There is a growing need to embed pluralistic value systems, including faith-informed and community-centric perspectives, into AI design and governance processes. Fifth, institutional capacity for implementation is a persistent gap. Many frameworks remain at the level of principles without translating into enforceable standards, audit mechanisms, or regulatory tooling. Governance effectiveness depends on executable models, not only normative alignment. Finally, geopolitical fragmentation and regulatory divergence pose risks to interoperability and global coordination. As different blocs pursue distinct AI strategies, there is a need for bridging mechanisms that allow cooperation without forcing uniformity. Addressing these cross-cutting issues would strengthen the dialogue's ability to move from high-level consensus toward equitable, actionable, and globally relevant AI governance.

Governance gaps in the selected areas are already shaping outcomes across Malaysia and the broader ASEAN region, particularly in public sector digitalisation, Islamic finance, and social impact ecosystems. A key challenge is the fragmentation between policy intent and implementation capacity. While national AI strategies and guidelines exist, institutional readiness, technical expertise, and regulatory tooling remain uneven. This affects the ability to operationalise safe, secure, and trustworthy AI, especially in high-impact domains such as financial services and public welfare systems. In terms of AI capacity-building, the region faces a structural talent gap. There is strong demand for data scientists, AI engineers, and governance specialists, but limited local pipelines. This creates dependency on external vendors and increases risks related to data control, system transparency, and long-term sustainability. The social and cultural implications of AI are particularly significant in a diverse, multi-ethnic, and multilingual context. Many AI systems are not adequately localised, leading to biases in language models, misalignment with local norms, and reduced effectiveness in community-facing applications such as digital philanthropy or citizen services. At the same time, open-source AI and open data ecosystems present a major opportunity. They enable governments, SMEs, and institutions to experiment, build, and deploy solutions at lower cost, accelerating innovation without heavy capital expenditure. This is especially relevant for zero-CAPEX partnership models and public-private collaborations. A critical emerging tension lies in balancing openness with governance. Without clear standards on data usage, model accountability, and security, open ecosystems can introduce new risks even as they expand access. Overall, the region stands at an inflection point: governance gaps create exposure to dependency, fragmentation, and misalignment, but if addressed strategically, they also open pathways for locally grounded, inclusive, and scalable AI ecosystems that align with national development priorities and societal values.

The AI Dialogue can play a pivotal role as a bridging platform that translates fragmented global efforts into coordinated, actionable cooperation. First, it can enable norm convergence without imposing uniformity. By facilitating structured exchanges between diverse jurisdictions, the Dialogue can help align core governance principles while respecting different legal systems, developmental stages, and cultural contexts. This is critical to reducing regulatory fragmentation and fostering mutual trust. Second, the Dialogue can act as a coordination mechanism for implementation, not just discussion. This includes establishing thematic working groups, shared policy toolkits, and reference architectures that countries can adapt locally. Moving from principles to deployable models would significantly accelerate adoption, especially for emerging economies. Third, it can strengthen multi-stakeholder collaboration at a global scale. Governments, industry, academia, and civil society each hold different pieces of the AI governance puzzle. The Dialogue can convene these actors into structured partnerships, such as cross-border pilot projects, regulatory sandboxes, and joint research initiatives. Fourth, the Dialogue can advance capacity-sharing and resource pooling. This includes technical assistance, knowledge exchange, and potentially shared infrastructure models. For many countries, access to expertise, data governance frameworks, and compute resources is as important as policy guidance. Fifth, it can serve as a platform for Global South leadership and inclusion. By ensuring equitable participation in agenda-setting and decision-making, the Dialogue can help shape governance models that are not solely driven by a few dominant actors. Finally, the Dialogue can institutionalise continuity and accountability, through periodic reporting, progress tracking, and iterative refinement of frameworks. In essence, its role is to move the global ecosystem from isolated initiatives toward collaborative, interoperable, and implementation-driven AI governance, anchored in shared responsibility and mutual benefit.

The AI Dialogue should build on existing global and regional initiatives while positioning itself as a convergence and implementation layer. Key multilateral efforts include the UNESCO Recommendation on the Ethics of AI, the OECD AI Principles, and the Global Digital Compact under the United Nations. These provide normative foundations on ethics, human rights, and responsible use. The Dialogue can align and operationalise these into deployable governance models. At the regulatory and standards level, frameworks such as the EU AI Act, ISO/IEC AI standards (e.g., ISO/IEC 42001), and the NIST AI Risk Management Framework offer concrete approaches to risk classification, compliance, and assurance. The Dialogue can help translate these into adaptable templates for countries with varying capacities. Multi-stakeholder platforms such as the Global Partnership on AI (GPAI) and the World Economic Forum AI Governance initiatives provide valuable research and policy experimentation. Regionally, ASEAN's Guide on AI Governance and Ethics is particularly relevant for emerging economies. The Dialogue can connect these ecosystems to avoid duplication and promote interoperability. In terms of added value, the AI Dialogue can: • Serve as a global coordination hub, linking fragmented initiatives into a coherent ecosystem • Provide implementation pathways, including toolkits, sandboxes, and pilot programmes • Enable capacity-sharing mechanisms, especially for Global South participation • Facilitate cross-border interoperability, reducing regulatory divergence • Promote contextualisation, ensuring frameworks are adaptable to local cultural and socio-economic realities Crucially, the Dialogue's unique contribution lies in shifting from principle-setting to execution at scale. By integrating existing standards, aligning stakeholders, and driving practical collaboration, it can accelerate the transition toward inclusive, interoperable, and action-oriented AI governance.

Effective participation in the AI Dialogue requires a structured, role-driven approach where each stakeholder contributes according to its comparative advantage. Governments should lead on policy direction, regulatory alignment, and international coordination. Their role is to define national priorities, enable regulatory sandboxes, and commit to pilot implementations. Industry should contribute technical expertise, real-world use cases, and scalable solutions. This includes sharing best practices on model governance, safety engineering, and deployment standards, while committing to responsible innovation. Academia and research institutions should provide independent validation, methodological rigor, and foresight on emerging risks. They play a critical role in benchmarking, auditing frameworks, and developing talent pipelines. Civil society and community organisations should ensure that societal values, inclusion, and rights-based perspectives are embedded. Their participation is essential to surface real-world impacts, especially among vulnerable or underrepresented groups. To be effective, the AI Dialogue should adopt a multi-layered structure: 1. Plenary Track: High-level alignment on principles, priorities, and geopolitical coordination 2. Thematic Working Groups: Focused tracks (e.g., safety, data governance, capacity-building) tasked with producing actionable outputs 3. Implementation Labs / Sandboxes: Pilot environments where policies and frameworks are tested in real-world scenarios 4. Regional Nodes: Decentralised platforms (e.g., ASEAN, Africa, EU) to contextualise global frameworks 5. Annual Reporting Mechanism: Tracking progress, commitments, and measurable outcomes In terms of format, the Dialogue should combine policy roundtables, technical workshops, and deployment-focused pilots, ensuring continuous engagement rather than one-off discussions. Ultimately, its effectiveness depends on moving from representation to co-creation and shared accountability, where stakeholders are not only consulted but actively responsible for delivering outcomes.

Several voices remain structurally underrepresented in global AI governance, limiting both legitimacy and effectiveness. First, the Global South is often included as participants but not as agenda-setters. Many developing countries face distinct challenges around data sovereignty, infrastructure, and socio-economic priorities, yet their perspectives are underweighted in shaping norms. Inclusion requires shifting from consultation to co-design, with equitable representation in decision-making bodies and leadership roles in working groups. Second, small and medium enterprises (SMEs) and local innovators are frequently overlooked. Governance discussions tend to be dominated by large technology firms, yet SMEs are critical to real-world deployment and local adaptation. Creating dedicated SME tracks, innovation sandboxes, and funding access can ensure their participation. Third, civil society organisations and grassroots communities, particularly those representing vulnerable populations, remain under-engaged. Their lived experiences are essential to understanding the societal impacts of AI, including issues of bias, access, and digital exclusion. Structured mechanisms such as community consultations, participatory design processes, and impact feedback loops should be embedded into the Dialogue. Fourth, non-Western epistemologies and value systems, including faith-informed, indigenous, and culturally grounded perspectives, are rarely operationalised in governance frameworks. This creates a risk of normative dominance by a narrow set of worldviews. Inclusion here requires deliberate space for alternative ethical frameworks and interdisciplinary contributions beyond purely technical or legal domains. Fifth, public sector practitioners from implementation agencies are often missing. While policymakers are present, those responsible for executing AI systems in healthcare, welfare, and education bring critical operational insight. To address these gaps, the AI Dialogue should institutionalise inclusive representation quotas, regional leadership nodes, capacity-building support, and funding mechanisms that enable meaningful participation. Broadening participation is not only a matter of equity but a prerequisite for context-aware, legitimate, and globally resilient AI governance.

To move beyond passive discussion, the AI Dialogue should adopt participatory and outcome-driven formats that blend policy, technical, and societal perspectives. First, Policy-to-Prototype Labs can translate governance ideas into working models. Mixed teams of policymakers, engineers, and domain experts co-develop practical outputs such as risk classification tools, audit templates, or data governance workflows. This ensures immediate applicability. Second, Scenario-Based Simulations ("AI Governance War Games") can stress-test policies against real-world situations, such as cross-border data disputes, model failures, or misinformation crises. These simulations reveal gaps in coordination, accountability, and response mechanisms. Third, Regulatory Sandboxes with Live Use Cases allow participants to test frameworks in controlled environments. Governments and industry can jointly pilot AI systems in sectors like finance, healthcare, or public services, generating evidence for scalable policy design. Fourth, Multi-Stakeholder Roundtables with Structured Outputs should replace open-ended panels. Each session should be tasked with producing a defined deliverable, such as a policy brief, standard draft, or implementation roadmap within a fixed timeframe. Fifth, Reverse Pitch Sessions can invert traditional formats. Instead of startups pitching solutions, governments and public institutions present real governance challenges, inviting industry and academia to propose solutions aligned with policy needs. Sixth, Regional Co-Creation Nodes can run in parallel, enabling context-specific discussions across ASEAN, Africa, Europe, and other regions, with outputs fed into the global plenary. This ensures localisation without fragmentation. Finally, a Digital Collaboration Platform should support continuous engagement before and after the event, enabling document co-creation, knowledge sharing, and progress tracking. These formats shift the Dialogue from static exchange to co-creation, experimentation, and implementation, ensuring that engagement produces tangible and scalable governance outcomes.

Several existing policies, practices, and platforms offer concrete pathways toward effective AI governance by combining principles with implementation. At the policy level, the EU AI Act provides a structured, risk-based approach that classifies AI systems by impact and imposes proportionate obligations. This model is valuable for translating abstract ethics into enforceable compliance. Similarly, the NIST AI Risk Management Framework offers a practical lifecycle approach, guiding organisations on identifying, measuring, and mitigating AI risks in real-world deployments. In standards and assurance, ISO/IEC 42001 (AI Management Systems) introduces auditable governance structures, enabling organisations to institutionalise accountability, documentation, and continuous improvement. This is critical for moving from voluntary guidelines to certifiable practices. From a regional perspective, the ASEAN Guide on AI Governance and Ethics demonstrates how principles can be contextualised for emerging economies, balancing innovation with safeguards while remaining adaptable to different national capacities. In practice, regulatory sandboxes have proven effective. Countries such as Singapore and the UK have used sandbox environments to test AI applications in controlled settings, allowing regulators and innovators to co-develop rules while reducing uncertainty and risk. On the platform side, emerging data governance infrastructures such as trusted data exchanges and federated data-sharing models enable collaboration without centralising sensitive data. These approaches address concerns around privacy, sovereignty, and cross-border interoperability. Additionally, open-source AI ecosystems provide transparency and accessibility, allowing independent scrutiny, local adaptation, and capacity-building. When combined with governance layers such as model documentation and audit trails, they can support responsible innovation at scale. Collectively, these examples show that effective AI governance requires an integrated approach: clear regulatory frameworks, operational standards, experimental environments, and enabling infrastructure, all aligned to ensure accountability while sustaining innovation.