Ministry of Economy and Finance

Honestly, I've been thinking about this a lot — not just as someone studying AI at the graduate level, but as someone who actually works inside a government ministry in Cambodia. Those two perspectives don't always line up neatly, and I think that tension is exactly what this Dialogue needs to sit with. The outcome I want most is simple: I want developing countries to leave Geneva feeling like this process was built for them too, not just handed down to them. In Cambodia and across ASEAN, we are not waiting for AI — it is already here, already being used in public finance systems, digital services, and government infrastructure. What is missing is not the technology. It is the governance thinking that should have come alongside it. A successful Dialogue begins by being honest about that reality. Second, I want real capacity building — not the kind that lives only in a report. I mean actual technical cooperation, peer learning between countries at similar stages, and support that reaches the people inside ministries who are making technology decisions every day. I am one of those people. I know what it feels like to move fast on infrastructure without a clear policy framework behind you. Third, I hope Geneva produces a structure that keeps this conversation alive — something that genuinely feeds into New York in 2027 with measurable progress, not recycled talking points. And finally, I want accountability treated as non-negotiable. When governments deploy AI in ways that affect citizens' lives — their finances, their access to services, their fundamental rights — there must be transparency and oversight. That cannot be a footnote. For me, that is the entire point of having this Dialogue in the first place. If those four things come out of Geneva, I would call it a success genuinely worth building on.

• AI capacity-building
• Safe, secure and trustworthy AI
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI

These four areas reflect what I live and work with every day, so my choices were not difficult to make. I selected AI capacity-building first because, working inside a government ministry in Cambodia, I see the gap between where we are and where we need to be. We have ambition. We have people who care. But the technical knowledge, the policy frameworks, and the institutional capacity to govern AI responsibly are still being built. Without serious investment in capacity building - real training, real cooperation, not just workshops - developing countries will keep being recipients of AI governance decisions made elsewhere, rather than active participants in shaping them. Safe, secure and trustworthy AI matters to me because of my day-to-day work as a DevOps Engineer. I deploy and maintain government infrastructure. I understand what it means when a system fails, when data is exposed, or when automation produces an outcome nobody expected. In a public sector context, those failures are not just technical problems - they affect real people and erode trust in government institutions. Safety and security are not optional extras. Transparency, accountability, and human oversight connects directly to that. As governments increasingly use AI in decision-making - in finance, in public services, in administration - citizens deserve to know how those decisions are made and to have recourse when things go wrong. I believe strongly that human oversight must remain central, especially in the early stages of AI adoption in government. Finally, the social, economic, ethical, cultural, linguistic and technical implications of AI matter deeply from a Cambodian and ASEAN perspective. AI systems built without considering local languages, cultural contexts, and economic realities will not serve our communities well. Governance frameworks that ignore these dimensions will always fall short. These four priorities, together, represent what I believe responsible AI governance must address.

Yes, and I think this is actually one of the more important questions in this form. The themes listed cover a lot of ground, but there is one area I feel is consistently underrepresented in global AI governance conversations: the reality of countries that are simultaneously building digital infrastructure and trying to govern AI at the same time. Most governance frameworks assume a baseline - that countries already have stable data infrastructure, functioning regulatory institutions, and a digitally literate workforce. For many nations in Southeast Asia, including Cambodia, that baseline is still being constructed. We are not governing AI after the fact. We are deploying it while still laying the foundational systems underneath it. That creates a very different set of risks and priorities that the listed themes do not fully capture. A second issue I would raise is AI and local language equity. Cambodia has its own language, script, and cultural context. The overwhelming majority of AI models - large language models, speech recognition systems, document processing tools - are built on English and a handful of dominant languages. When governments in countries like ours try to use these tools in public services, the performance gap is real and significant. This is not just a technical problem. It is a governance and equity problem, and it deserves its own dedicated attention rather than being absorbed into a broader theme. Third, I think the intersection of AI and sovereign data governance needs more explicit treatment. As developing countries digitize their public services, questions about who owns government data, where it is stored, and how AI vendors access it become critically important. These are not purely technical questions - they are sovereignty questions. These issues cut across all the listed themes but do not sit cleanly inside any one of them. They deserve to be named directly.

I want to answer this honestly, from the ground level, because I think that is where the most useful perspective comes from. Cambodia is in an interesting and sometimes uncomfortable position. We are a developing country that is genuinely committed to digital transformation. The Ministry of Economy and Finance, where I work, has been actively building and deploying digital public finance systems — e-invoicing platforms, digital authentication infrastructure, government data pipelines. The ambition is real. But the governance frameworks that should sit alongside that ambition are still catching up. The most significant challenge I see is the capacity gap. We have engineers and technical staff working hard to build and maintain these systems, but the policy and regulatory expertise needed to govern AI responsibly is thin. There are very few people in government who understand both the technical realities of AI and the policy implications deeply enough to bridge that gap. I am studying for my master's degree in Data Science and Engineering precisely because I feel that gap personally and want to help close it. The second challenge is dependency. Much of the AI and cloud infrastructure that developing countries rely on is built, owned, and governed by foreign companies operating under foreign regulations. That creates real vulnerabilities — in data sovereignty, in service continuity, and in the ability to hold systems accountable when things go wrong. But there are genuine opportunities too. Cambodia and the broader ASEAN region are young, growing, and increasingly connected. We have the chance to build governance frameworks that are fit for our context from the beginning, rather than retrofitting regulations designed elsewhere. The AI Dialogue is exactly the kind of platform that could support that — if it takes developing country realities seriously and moves beyond declarations into practical cooperation. That is the opportunity I am hoping this Dialogue will not waste.

I think the AI Dialogue has a role that no other existing forum quite fills — but only if it is willing to be honest about what international cooperation actually requires. Right now, global AI governance conversations tend to happen in spaces dominated by a handful of technologically advanced nations and large private sector actors. The outcomes of those conversations — standards, norms, frameworks — then flow outward to the rest of the world, often without meaningful input from the countries that will be most affected by them. The AI Dialogue, sitting within the United Nations system, has a genuine opportunity to change that dynamic. But it has to be intentional about it. The most valuable role the Dialogue can play is creating a structured, trusted space where countries at very different stages of AI development can speak honestly about what they are experiencing — the pressures, the gaps, the dependencies, the risks — without feeling like they are falling behind or being judged against a standard they had no hand in setting. Beyond conversation, I believe the Dialogue should actively facilitate peer-to-peer cooperation between countries. Not just North-South technical assistance, but genuine South-South exchange. Countries like Cambodia can learn enormously from others in ASEAN, Africa, and Latin America who are navigating similar challenges with similar resources. That kind of cooperation does not happen automatically. It needs a platform and a mandate to make it happen. The Dialogue can also play a critical role in translating global principles into nationally applicable guidance. High-level declarations are easy to agree on. What is hard — and what actually matters — is helping governments understand what those principles mean for their specific legal systems, their specific institutions, and their specific technical realities. That translation work is where I believe the Dialogue can make its most lasting contribution.

There are already meaningful initiatives out there, and I think one of the smartest things the AI Dialogue can do is resist the temptation to start from scratch. The world does not need another framework that ignores everything that came before it. A few existing efforts stand out as worth building on directly. The UNESCO Recommendation on the Ethics of AI is one of the most inclusive normative frameworks developed so far, with genuine developing country participation in its drafting. The AI Dialogue should treat it as a foundation, not a competitor. Similarly, the ITU's AI for Good platform has built real connections between technical communities and policy conversations — that network and credibility should be leveraged, not duplicated. Within my own region, ASEAN's Guide on AI Governance and Ethics represents a serious attempt to develop regionally contextual principles. The AI Dialogue should engage with regional frameworks like this one actively, because they carry legitimacy that a purely global framework sometimes lacks at the national level. The OECD AI Principles have also shaped a lot of national AI strategies, including in countries that are not OECD members. The Dialogue should find ways to connect with that work while ensuring it does not become a mechanism that simply exports OECD-centric thinking to the rest of the world under a UN banner. What the AI Dialogue can add that none of these initiatives fully provide is universal legitimacy combined with operational support. The UN system reaches governments that other forums do not. If the Dialogue uses that reach to connect existing initiatives, broker cooperation between them, and direct resources toward implementation in developing countries — rather than producing yet another standalone declaration — it will have added something genuinely valuable. That connective role is where I believe the Dialogue's unique contribution lies.

I want to start with something that I think gets overlooked in discussions about format and structure: who is actually in the room matters more than how the room is arranged. Too many international dialogues on technology governance end up being conversations between the same set of voices — large member states, established research institutions, and well-resourced private sector actors. Civil society organizations from developing countries, young technical professionals working inside government ministries, researchers from universities that are not in the global north — these perspectives are rarely centered. If the AI Dialogue is serious about being inclusive, it needs to design for inclusion deliberately, not just declare it. In terms of format, I would strongly recommend moving away from plenary-heavy structures where a handful of delegations speak at length while everyone else listens. The sessions that actually produce useful outcomes in my experience are smaller, more focused working groups where people with different backgrounds — technical, legal, policy, civil society — work through specific problems together. That kind of structured dialogue produces something that large formal sessions rarely do: genuine mutual understanding. I would also recommend building in dedicated time and space for developing country governments to share their experiences directly — not as case studies presented by outside researchers, but in their own words, on their own terms. There is a lot of practical knowledge sitting inside ministries like the one I work in that never makes it into global policy conversations because the format does not create space for it. For ongoing engagement between sessions, a simple and accessible digital platform — one that works reliably in lower-bandwidth environments — would make a real difference. Not every stakeholder can travel to Geneva or New York. The Dialogue should not require them to in order to participate meaningfully. Structure should serve inclusion. That is the principle I would keep coming back to.

This question feels personal to me, because in many ways I am one of those underrepresented voices. I am a young technical professional from Cambodia, working inside a government ministry, studying AI at the graduate level. I follow global AI governance conversations closely. I read the frameworks, the declarations, the policy papers. And yet the experiences I live — deploying government infrastructure in a developing country context, navigating the gap between ambitious digital transformation goals and limited institutional capacity, working with AI tools that were not built with my language or my context in mind — almost never appear in those conversations. Not because they are unimportant, but because the spaces where these discussions happen were not designed with people like me in mind. The most underrepresented voices I would highlight are these: government technical staff in developing countries who understand both the technology and the public sector reality but rarely get invited to international forums. Researchers and academics from the Global South whose work on locally relevant AI challenges does not get the same visibility as research coming out of institutions in wealthy nations. Indigenous and minority language communities whose languages are almost entirely absent from major AI systems and whose cultural contexts are rarely considered in ethics discussions. And young professionals and students who will live with the consequences of today's governance decisions longest but have the least formal influence over them. Inclusion is not just about sending invitations. It requires covering participation costs for those who cannot self-fund, providing interpretation and translation so language is not a barrier, designing consultation processes that work across time zones and bandwidth limitations, and genuinely incorporating the inputs received rather than treating them as a formality. The test of inclusion is simple: do the outcomes of this Dialogue reflect the perspectives of those who are most affected by AI but least able to shape it? If yes, it worked. If not, it did not.

I think the honest answer is that the most innovative thing the AI Dialogue could do is abandon the formats that international forums have relied on for decades and actually experiment with something different. Traditional conference structures — keynote speeches, panel discussions, formal plenary sessions — are comfortable and familiar, but they are not where real understanding gets built. In my experience, both in government work and in academic settings, the conversations that actually shift thinking happen in smaller, less formal spaces where people feel safe enough to say what they genuinely believe rather than what their institution has approved. With that in mind, a few formats I think could make a real difference. First, structured problem-solving sessions where mixed groups — a government official, a researcher, a civil society representative, a technical practitioner — are given a specific, real-world AI governance challenge and asked to work through it together. Not to produce a declaration, but to actually think out loud as a group. That kind of format surfaces practical knowledge that polished presentations never do. Second, live cross-regional exchanges where government teams from different countries share what they are actually building and governing right now — the systems, the challenges, the mistakes, the lessons. Peer learning of that kind is far more useful to a ministry like the one I work in than another set of high-level principles. Third, for those who cannot attend in person, asynchronous contribution mechanisms that go beyond a simple online form. Facilitated virtual working groups, recorded input sessions with real response and follow-up, multilingual discussion threads — these would make remote participation feel meaningful rather than performative. And finally, I would love to see space reserved specifically for younger professionals and students to present their perspectives directly — not as a side event, but as a core part of the programme. Format shapes who speaks and who is heard. It is worth getting right.

I want to share examples that feel real and grounded rather than just listing the usual suspects that appear in every AI governance paper. The first example I keep coming back to is Singapore's Model AI Governance Framework. What makes it stand out is not that it is perfect, but that it was designed to be practical. It gives organizations - including government agencies - concrete, actionable guidance rather than abstract principles. For countries in ASEAN that are trying to move from policy intention to actual implementation, that kind of practical framework is far more useful than a beautifully written declaration that nobody knows how to apply. The second example is Estonia's approach to digital governance broadly. Estonia built its entire digital public infrastructure on principles of transparency, interoperability, and citizen control over personal data long before AI was the dominant conversation. The lesson I take from Estonia is not to copy their specific solutions - our contexts are too different - but to recognize that getting the foundational digital governance right creates the conditions for responsible AI adoption later. Cambodia is still working on those foundations, and that sequencing matters. Third, within my own work context, I have seen the value of incremental, monitored deployment of digital systems in government. Rather than launching fully automated systems immediately, building in human review stages, feedback mechanisms, and clear escalation paths has helped catch problems early and maintain public trust. It is not glamorous, but it works. Finally, I think UNESCO's ethical impact assessment methodology for AI systems deserves more attention as a practical tool. It gives governments and institutions a structured way to think through the implications of an AI system before deploying it, which is exactly the kind of upstream governance thinking that prevents problems rather than just responding to them. These are not perfect models. But they are honest ones.