Joowal

The first Global Dialogue on AI Governance will be a genuine success only if it produces outcomes that are structurally irreversible, not merely declaratory. Three concrete outcomes would meet that standard. First, the Dialogue must establish a binding commitment to redistribute AI governance agenda-setting power to the Global South. As a startup founder building in Southeast Asia, I observe daily that the governance frameworks shaping my operating environment are designed in Brussels, Washington, and Beijing, and then exported to regions like mine as finished products. Many Global South governments have emphasised sovereignty in AI governance and the aspiration to be active partners rather than norm-takers. The Geneva session must translate that aspiration into institutional reality by creating formal co-drafting mechanisms that give developing-country stakeholders, including founders and entrepreneurs, binding input into the norms that will govern their industries. Second, the Dialogue must produce a concrete framework for compute and data sovereignty. The UN's own Special Envoy for Digital and Emerging Technologies has identified the concentration of economic and technological power as the organization's greatest concern, warning that countries which miss the current AI wave risk falling fifty years behind in development, echoing the consequences of missed industrialisation. Sovereignty over AI infrastructure is not a peripheral concern but a prerequisite for meaningful participation in the global economy. Third, the Dialogue must mandate a structured accountability mechanism for South-South AI cooperation, giving regions like ASEAN a recognised pathway to develop interoperable governance frameworks without requiring adoption of frameworks designed for entirely different political economies. Governance tracks enabling regulators from the Global South to jointly test proportionate, risk-based approaches in sandboxed settings would move the international community beyond consensus governance toward practical interoperability.

• AI capacity-building
• Open-source software, open data and open AI models
• Interoperability of governance approaches
• Social, economic, ethical, cultural, linguistic and technical implications of AI

As a startup founder operating in Southeast Asia, my four selections form a single, interlocking argument about what AI sovereignty means in practice for the Global South, and why the current trajectory of global AI governance is failing to address it. The social, economic, ethical, cultural, and linguistic implications of AI represent the foundational priority because AI systems are not culturally neutral infrastructure. They embed the values, languages, and economic assumptions of whoever builds them. Founders in Southeast Asia are deploying models trained predominantly on Global North data into societies with profoundly different linguistic diversity, economic structures, and ethical frameworks. The downstream harm of that mismatch is already measurable and will compound without deliberate intervention. Interoperability of governance approaches is the most operationally urgent priority for any founder building across ASEAN. The current landscape of voluntary, fragmented, and mutually unrecognised national frameworks does not create regulatory clarity. It creates a vacuum that defaults to whichever AI ecosystem imposes the least compliance friction, a dynamic that structurally disadvantages locally built, values-aligned alternatives before they can compete on merit. AI capacity-building is sovereignty made concrete. Without indigenous compute infrastructure, locally trained models, and technical literacy distributed across the innovation ecosystem, every governance framework a developing country adopts remains dependent on infrastructure it does not control. Principles without capability are not governance but mere aspiration. The open-source dimension is perhaps the most consequential and least scrutinised priority. The rapid rise of Chinese open-source large language models, which accounted for approximately thirty percent of global usage share by the end of 2025, means that the de facto AI stack across much of Southeast Asia is being set by actors whose governance philosophies are embedded at the model level. Who governs open-source norms governs the default technological choices of the Global South.

There is one cross-cutting issue of profound urgency that none of the listed themes adequately captures, namely the weaponisation of AI infrastructure as a geopolitical instrument, and the consequences for Global South nations that host the physical substrate of this competition without governing it. The listed themes treat AI governance as primarily a normative challenge, concerning rights, transparency, and capacity. What they collectively miss is that the infrastructure layer itself has become a site of active geopolitical contest shaping every other governance question. A November 2025 investigation revealed that a Shanghai-based startup, INF Tech was remotely accessing approximately 2,300 Nvidia Blackwell chips through an Indonesian data center, while Chinese technology firms including Alibaba and ByteDance were similarly using chips housed in Southeast Asian facilities to train frontier large language models, circumventing US export controls. Southeast Asian countries hosting this infrastructure had no governance framework adequate to their position in this dynamic. This is not merely a trade or security matter but a governance gap placing countries like Indonesia, Malaysia, and Thailand in an impossible structural position where their territory is the battlefield for great-power AI competition, their regulatory frameworks are insufficient to manage the consequences, and neither the US-led nor China-led governance architectures were designed with their interests as the primary concern. Analysis published in early 2026 observed that the global AI ecosystem is fracturing into competing technology stacks, and countries across Southeast Asia are each attempting to build partially sovereign digital infrastructure without a neutral multilateral framework to anchor that effort. The AI Dialogue is uniquely positioned to address this gap precisely because it sits above the bilateral contest. An explicit theme on AI infrastructure neutrality, covering compute access, data center governance, and third-country obligations, would give the Dialogue its most distinctive and consequential contribution to global governance.

The governance gaps in my selected thematic areas manifest not as abstract policy problems but as daily operational constraints with compounding consequences for the region's AI sovereignty. The most immediate challenge is what I describe as regulatory arbitrage by default. Across ASEAN, AI governance frameworks remain almost entirely voluntary and non-binding, with Vietnam only becoming the first member state to enact a formal AI law in December 2025. In this vacuum, founders building across multiple markets face irreconcilable compliance signals. The practical result is that product decisions are driven not by values-aligned governance but by whichever regulatory environment imposes the least friction. This structurally advantages large foreign platforms over locally built alternatives, compounding dependency rather than reducing it. The open-source dimension makes this worse in ways the thematic list does not adequately surface. Chinese open-source large language models accounted for approximately thirty percent of global usage share by end-2025, up from thirteen percent at the start of that year, with Malaysia announcing that its sovereign AI ecosystem would be built on DeepSeek. This is not merely a technology procurement decision but a governance decision with long-term implications for which values, linguistic assumptions, and accountability architectures become embedded in the region's AI infrastructure. The opportunity, however, is equally significant and equally time-sensitive. Stanford University's 2026 AI Index Report documents that Southeast Asian publics express the highest levels of AI optimism and government trust in the world, with Singapore at eighty-one percent, Indonesia at seventy-six percent, and Malaysia at seventy-three percent trusting their governments to regulate AI responsibly. This trust surplus is a governance asset that no Western democracy currently possesses. It creates a narrow but genuine window to build legitimate, participatory AI governance institutions before deployment outpaces accountability and that trust erodes irreversibly.

The AI Dialogue's most distinctive and irreplaceable contribution to international cooperation is one that no bilateral agreement, regional framework, or industry-led initiative can replicate as it is the only forum where the geopolitical contest over AI governance must yield, at least procedurally, to universal participation. That structural fact is its greatest asset and the source of its most consequential potential role. From the perspective of a founder operating in Southeast Asia, the Dialogue's most urgent cooperative function is to serve as a neutrality anchor for regions caught between competing AI stacks. The Atlantic Council observed in early 2026 that the global AI ecosystem is fracturing into opposing infrastructure approaches, with the United States explicitly exporting its technology stack as stated policy and China doing the same through open-source models and data center partnerships across the region. Countries in Southeast Asia are not passive observers of this contest but act as the terrain. The Dialogue is the only forum with the legitimacy and universality to establish principles of AI infrastructure neutrality that neither Washington nor Beijing can unilaterally impose or veto. Second, the Dialogue can play an indispensable role in converting national governance experiments into internationally recognised standards. Vietnam enacted the region's first formal AI law in December 2025, Indonesia is finalising presidential regulations, and Singapore has produced the world's first AI governance testing toolkit. These are genuinely valuable innovations, but they are currently isolated. The Dialogue can function as a mutual recognition mechanism, giving regional governance efforts global standing without forcing homogenisation onto jurisdictions with vastly different institutional capacities. Third, and most consequentially, the Dialogue must become the venue where founders, practitioners, and civil society from the Global South are not merely consulted but structurally integrated into the drafting of norms.

Malaysia's National AI Office (NAIO - launched in December 2024), represents one of the most operationally serious attempts by a middle-income Global South country to build a centralized, cross-sector AI governance institution. Its AIForMYFuture initiative had already skilled more than 400,000 Malaysians by May 2025, and its AI Technology Action Plan 2026 to 2030 provides a replicable model for national capacity-building that other developing countries could adapt. The AI Dialogue should treat NAIO not merely as a participant but as a blueprint contributor, using its experience to inform a standard framework for national AI offices in developing economies. The ASEAN Working Group on AI Governance, chaired by Singapore's IMDA, and the ASEAN AI Safety Network whose secretariat was announced to be based in Kuala Lumpur in early 2026, represent the most substantive regional governance architecture currently active in the Global South. The AI Dialogue should formally recognise these structures as regional nodes, giving their outputs standing in the global architecture without requiring them to abandon contextual specificity. Women in AI Governance and UNESCO's Women4Ethical AI initiative address a structural deficit that every other mechanism underserves. Women remain underrepresented across the entire AI lifecycle from data design to deployment, and Southeast Asia is no exception despite its high digital adoption rates. The AI Dialogue should mandate gender-disaggregated participation requirements across all working groups, treating inclusion not as a courtesy but as a validity condition for any governance output it produces.

The Dialogue must move decisively beyond traditional UN plenary formats. Intergovernmental statements read into the record by career diplomats do not produce governance innovation. The European Union's April 2026 submission to the Co-Chairs explicitly called for participatory and interactive sessions focused on practical exchange rather than formal positioning, and this recommendation deserves implementation. Governments should contribute by mandating that their delegations include actual practitioners. Industry should contribute by disclosing governance architectures, not only aspirational principles. Civil society, and specifically founders and entrepreneurs from the Global South, should contribute by presenting operational evidence of how governance gaps affect real products and real users. The Dialogue also needs a permanent practitioner track that runs parallel to the intergovernmental track and feeds directly into it. Informal convenings ahead of the Global Digital Compact negotiations demonstrated that bringing diplomats, technical experts, and civil society into the same room, outside formal procedures, grounded negotiations in lived reality in ways that plenary sessions cannot replicate. That model should be institutionalised and not treated as a preparatory courtesy. Finally, the Dialogue should adopt a pre-session submission architecture modelled on the Universal Periodic Review process, where each stakeholder group submits a structured account of AI governance conditions in their sector or region before each session. This would generate a comparative evidence base that the Independent International Scientific Panel could draw upon, creating a feedback loop between scientific assessment and stakeholder experience that neither body can currently sustain alone.

Founders from developing economies interact daily with governance gaps that produce harm, yet hold no formal standing in processes that could close them. Global forums are dominated by large-platform representatives whose interests reflect incumbent scale, not the conditions facing a startup in Kuala Lumpur or Jakarta. Indigenous and linguistically marginalized communities face compounded exclusion. Research in 2026 confirmed that AI foundation models encode systematic bias against Global South contexts, producing what researchers called geographical hallucination, where models generate confidently wrong representations of developing-country realities. Indigenous peoples additionally face extractive appropriation of their language data without consent, while remaining absent from forums where such practices could be regulated. Data laborers, the moderators and annotators across the Philippines, Kenya, and Indonesia whose work sustains global AI systems, remain invisible in every governance room. Inclusion also requires funded participation pathways, interpretation in non-dominant languages, asynchronous mechanisms for remote engagement, and formal standing for non-state actors in working group outputs. The AI Dialogue has travel support provisions for developing-country participants. The test is whether that architecture will reach communities most harmed, not only governments that claim to represent them.

I would suggest a structured red-team track where practitioners from developing economies publicly stress-test proposed governance frameworks against real deployment conditions, producing documented failure cases that inform working group outputs rather than disappearing into plenary records. A binding pre-session evidence protocol requiring each stakeholder group to submit quantified governance gap assessments before each session, creating a comparative baseline the Independent Scientific Panel can interrogate, may also be useful to consider. Lastly, I would recommend asynchronous deliberation windows of four weeks before each session, with translation into non-dominant languages, ensuring participation is not rationed by travel budgets or time zones.

The following four approaches constitute a Southeast and South Asian governance toolkit that the AI Dialogue should formalise as a Global South reference architecture: Malaysia's NAIO demonstrated that national AI capacity-building can be operationalized at scale through structured public-private partnerships, skilling more than 400,000 citizens within six months of its AIForMYFuture initiative's launch. Singapore's AI Verify toolkit, jointly developed by IMDA and the Personal Data Protection Commission, enables organisations to assess AI systems against governance principles through verifiable process checks rather than self-declared compliance. In January 2026, Singapore extended this approach with the world's first Model AI Governance Framework for Agentic AI, published at Davos, demonstrating that governance frameworks can keep pace with frontier deployment if institutional will exists. The AI Dialogue should treat AI Verify as a foundation for a global mutual recognition standard rather than allowing it to remain a national tool. India's IndiaAI Mission deployed a publicly funded common compute pool available to startups and researchers at subsidized rates well below global market prices, directly addressing the infrastructure dependency that makes AI sovereignty impossible for most developing nations. This model of compute-as-public-infrastructure is the most replicable governance innovation the Global South has produced. Vietnam's AI Law, enacted in December 2025, provides the region's first example of a developing country enacting binding AI legislation with phased compliance timelines and sector-specific risk classification. Its approach of delegating high-risk determinations to sectoral bodies reflects a governance model appropriate for jurisdictions with diverse institutional capacities.