Dreamers-Media Pacific

A successful first Global Dialogue on AI Governance should move beyond broad principles and deliver practical, implementable outcomes that reflect diverse global realities especially those of underrepresented regions. First, success would mean establishing a shared baseline for responsible AI governance not as a rigid global standard, but as a set of interoperable principles that countries can adapt. This includes alignment on transparency, accountability, data protection, and human oversight, with clear pathways for implementation in both high-capacity and resource-constrained contexts. Second, the Dialogue should produce actionable mechanisms for collaboration. This includes commitments to open standards, shared technical infrastructure, and knowledge exchange platforms that enable governments, civil society, and technical communities to co-develop solutions. For regions like the Pacific, where capacity varies, access to such shared systems is critical to avoid exclusion from the AI ecosystem. Third, a key outcome would be capacity-building commitments, funding, training, and institutional support to ensure that all countries can meaningfully participate in AI governance. This must include investment in local expertise, regulatory capability, and digital infrastructure, not just policy frameworks. Fourth, success would involve embedding trust and legitimacy into AI governance processes. This requires inclusive representation, ensuring voices from the Global South, Indigenous communities, and smaller states are not only present but influential in shaping decisions. Governance must reflect lived realities, not just technical or geopolitical priorities. Finally, the Dialogue should result in clear next steps and accountability structures such as working groups, timelines, and measurable outcomes so that commitments translate into sustained action. Drawing on experience delivering data systems, governance frameworks, and AI-enabled platforms across governments and regional organisations , meaningful impact comes not from declarations alone, but from systems that can be built, integrated, adopted, and sustained.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models

My selections reflect what I have seen working across governments and regional organisations in the Pacific, AI governance only works when it is practical, trusted, and adaptable to real-world constraints. Safe, secure and trustworthy AI is critical, particularly in environments where systems support public services, education, and national infrastructure. In my work building knowledge platforms and data systems across the region, trust is not abstract if systems are not secure and reliable, they are simply not used. Transparency, accountability, and human oversight are what make that trust operational. Many governments I've worked with require clear ownership of decisions, auditability of systems, and the ability to intervene. Governance must therefore translate into systems that are explainable and aligned with existing institutional processes. The social, cultural, and linguistic implications of AI are especially important in the Pacific. Systems must reflect local realities whether that is language, cultural knowledge, or community structures. Through co-design workshops and platform development, I've seen that adoption depends on whether people see themselves and their context reflected in the system. Finally, open-source software, open data, and open AI models are essential for equity. Many Pacific countries cannot rely solely on proprietary systems. Open approaches enable adaptation, local ownership, and long-term sustainability, while also supporting regional collaboration. These priorities reflect a shift from high-level principles to systems that can actually be built, governed, and sustained in diverse contexts like the Pacific.

Yes, there are several cross-cutting issues that are not fully captured, particularly when viewed through the lens of delivering real systems across the Pacific. First is implementation realism. Many governance discussions remain at the level of principles, but the challenge is translating these into systems that can operate within constrained infrastructure, limited technical capacity, and evolving institutional processes. In my experience building knowledge platforms and data systems for governments and regional organisations, the gap is not intent, it is execution. Second is data governance and sovereignty in practice. While human rights and ethics are covered, there is less emphasis on how data is actually managed across borders, institutions, and platforms. In the Pacific, questions of data ownership, control, and stewardship especially for cultural and community knowledge are critical and require operational models, not just policy statements. Third is sustainability of AI systems over time. Many solutions are deployed through short-term projects, but there is insufficient focus on maintenance, local ownership, and long-term financing. Systems that are not designed for continuity quickly become obsolete, regardless of how well they align with governance principles. Fourth is integration with existing systems and workflows. AI does not operate in isolation. It must integrate with legacy government systems, data pipelines, and institutional processes. Without interoperability at a practical level, even well-governed AI systems fail to deliver impact. Finally, there is an emerging need for community-centred co-design as a governance mechanism. In my work facilitating workshops and building platforms, governance is strongest when communities are directly involved in shaping how systems function, not just consulted after the fact. These issues highlight that effective AI governance must move beyond frameworks toward implementation, sustainability, and local ownership.

Governance gaps in AI are already shaping how technology is adopted across the Pacific, particularly in government, education, and regional coordination systems. The most significant challenge is the gap between policy ambition and implementation capacity. Many countries are engaging in AI and digital strategies, but lack the institutional structures, technical capability, and infrastructure to operationalise them. In my work delivering knowledge platforms and data systems across the region, this often results in fragmented solutions or over-reliance on external vendors, which can limit long-term ownership and sustainability. A second challenge is data governance and sovereignty in practice. Pacific countries are increasingly generating and relying on data, but there are limited frameworks for managing data access, sharing, and protection, especially across borders. This creates risks around control of sensitive data, including cultural knowledge and national datasets, and limits the ability to safely adopt AI systems. There is also a growing issue of system interoperability and integration. Governments often operate multiple legacy systems that are not designed to work together. Without clear governance on standards and integration, AI solutions struggle to deliver meaningful impact because they cannot access or align with existing data and workflows. However, these gaps also present clear opportunities. There is an opportunity to build fit-for-purpose governance models that are designed for smaller states, modular, adaptable, and grounded in real operational contexts. The Pacific can lead in demonstrating how AI governance can be embedded directly into system design, rather than treated as a separate policy layer. There is also strong potential in open and collaborative approaches, including open-source platforms and regional knowledge-sharing systems, which can reduce duplication and strengthen collective capability. Ultimately, the region has an opportunity to shape AI governance in a way that prioritises trust, sovereignty, and long-term sustainability from the outset.

The AI Dialogue can play a critical role by shifting international cooperation from high-level alignment to practical coordination and shared implementation. First, it can act as a convening platform that bridges policy, technical, and operational communities. Too often, AI governance discussions sit at the policy level, disconnected from those building and deploying systems. The Dialogue can bring these groups together to ensure that governance frameworks are informed by real-world system design, constraints, and use cases. Second, it can enable interoperability across governance approaches. Countries will continue to develop their own regulatory and policy models, but the Dialogue can help define common baselines, standards for data governance, auditability, risk classification, and accountability, that allow systems to work across borders. This is particularly important for regions like the Pacific, where collaboration across countries and institutions is essential. Third, the Dialogue can drive collective capacity-building. Many countries face similar challenges in technical capability, infrastructure, and institutional readiness. Through shared resources, training initiatives, and open technical frameworks, the Dialogue can reduce duplication and accelerate adoption in lower-capacity environments. Fourth, it can support open and collaborative ecosystems. Promoting open-source tools, shared datasets (with appropriate safeguards), and reusable system components enables countries to build on each other's work rather than starting from scratch. In my experience building regional platforms, this approach is key to sustainability and scale. Finally, the Dialogue should establish clear mechanisms for ongoing collaboration, working groups, pilot projects, and regional partnerships so that cooperation continues beyond the event itself. In this way, the AI Dialogue can move from discussion to delivery, enabling countries to not only align on principles, but to co-build systems that are trusted, interoperable, and sustainable.

The AI Dialogue should build on what is already working, but more importantly, connect and operationalise it. There are strong global foundations in place, UN digital cooperation efforts, OECD AI Principles, and the growing ecosystem around Digital Public Goods and open-source platforms. These provide the right direction. At the same time, development partners and regional organisations are already investing heavily in data systems, AI pilots, and digital infrastructure. From my experience working across the Pacific particularly building knowledge platforms and supporting regional systems through organisations like SPC the challenge is not a lack of initiatives. It is that they are often fragmented, not interoperable, and not designed for long-term ownership. This is where the AI Dialogue can add real value. First, it can connect policy to implementation. Many frameworks exist, but there is a gap in translating them into systems that governments can actually deploy, manage, and sustain. Second, it can push for interoperability in practice, shared standards, modular architectures, and reusable components that allow countries and organisations to build on top of each other's work, rather than duplicating effort. Third, it can strengthen regional collaboration models. In the Pacific, there is a real opportunity to share infrastructure, knowledge, and governance approaches across countries, but this needs intentional coordination and support. Fourth, it can anchor open approaches as the default, open-source, open standards, and where appropriate, shared datasets so that smaller states are not locked out of AI development. Ultimately, the added value of the Dialogue is moving from conversations to co-building systems, ones that are trusted, locally owned, and designed to work in the environments they are deployed in.

For the AI Dialogue to be effective, it needs to move beyond a traditional conference format and be structured as a working forum that produces tangible outputs. Different stakeholders should contribute based on their role in the ecosystem: Governments should bring real policy challenges and use cases, what they are trying to regulate, deploy, or manage. This grounds the Dialogue in reality rather than theory. Technical practitioners and builders should contribute system-level insights, what is actually feasible, what breaks in implementation, and how governance can be embedded into architecture. From my experience building platforms in the Pacific, this perspective is often missing but critical. Civil society and communities should shape how AI impacts people, ensuring cultural, ethical, and social realities are reflected in both policy and system design. Private sector and industry should contribute innovation pathways, but also align on responsibility, interoperability, and openness where possible. Development partners and academia can support with research, funding models, and capacity-building frameworks. In terms of format and structure, the Dialogue should be designed around delivery, not just discussion: 1. Problem-led working sessions: Small, mixed groups working on real challenges (e.g. data governance, AI in education, cross-border systems), with clear outputs. 2. System demonstrations and case studies: Show what has actually been built, what worked, what didn't, and why. This creates practical learning. 3. Co-design labs: Structured sessions where policy and technical actors jointly design governance approaches or system architectures. 4. Regional tracks: Dedicated space for regions like the Pacific to address shared constraints and opportunities. 5. Clear outputs and follow-through: Each track should produce actionable outputs, guidelines, pilot concepts, or working groups with defined next steps. The Dialogue should ultimately function as a platform for co-building governance in practice, not just aligning on principles.

Several voices remain consistently underrepresented in global AI governance discussions, particularly when viewed from the Pacific and similar contexts. First are small island states and low-capacity governments. These countries are often policy takers rather than shapers, yet they face some of the most immediate impacts of digital transformation. In my work across the Pacific, I've seen that constraints around infrastructure, funding, and technical capacity are rarely reflected in global frameworks. Second are technical practitioners working in public sector delivery the people actually building and maintaining systems. Governance is often discussed at a policy level, but those translating policy into real platforms, data systems, and workflows are not always at the table. This creates a disconnect between what is proposed and what is implementable. Third are local communities, including Indigenous groups, whose knowledge systems, languages, and data are increasingly intersecting with AI. Their perspectives are critical, particularly around data sovereignty, consent, and cultural integrity, yet they are often included only in consultation, not in decision-making. Fourth are educators and frontline service providers. In sectors like education, where I've worked on systems and emerging AI use cases, teachers and practitioners are directly affected by AI but rarely shape how it is governed. To include these voices, the Dialogue should move beyond open invitations and take a more intentional design approach: Dedicated regional representation with funded participation, particularly for small states. Practitioner-led sessions where builders and implementers present real systems and challenges. Community co-design mechanisms, not just consultation panels. Partnerships with regional organisations to aggregate and bring forward collective perspectives. Inclusion needs to be structured, resourced, and embedded into decision-making otherwise, the same gaps will persist.

The AI Dialogue needs to move away from passive panels and into working formats that actually produce outcomes. The most effective approach is problem-driven co-design labs small groups of policymakers, engineers, and community reps working on real challenges and producing tangible outputs. In my experience running workshops in the Pacific, this is where alignment actually happens. Second, we need live system demos and honest failure reviews. Not just what worked, but what didn't and why. That's where the real learning is. Third, regional solution sprints are important bringing together countries with similar constraints to co-develop shared approaches, whether that's governance templates or open-source tools. Fourth, policy–technical translation sessions can help bridge the gap between ambition and implementation by stress-testing ideas in real time. Finally, the Dialogue should include open build spaces where people can actively prototype, adapt, or explore tools during the event not just talk about them. Most importantly, all of this needs to lead somewhere. These sessions should feed into ongoing working groups, pilots, or partnerships so the Dialogue becomes a starting point for co-building, not just a one-off conversation.

Effective AI governance is emerging through a combination of policy, community-led frameworks, and system design, especially when grounded in local context. In the Pacific, a practical approach is governance-by-design embedding controls directly into platforms (e.g. role-based access, data ownership, and classification). In my experience building regional systems, this ensures governance is applied in practice, not just policy. In Aotearoa New Zealand, the government's Algorithm Charter sets clear expectations for transparency, accountability, and human oversight in public sector AI. This is complemented by Māori leadership through Te Mana Raraunga, which advances Māori Data Sovereignty, ensuring data is governed according to Indigenous rights, values, and collective ownership. Similarly, First Nations data sovereignty movements globally including frameworks like the CARE Principles emphasise authority to control, collective benefit, and ethical stewardship of data. These approaches shift governance from extractive models to ones grounded in community rights and long-term wellbeing. There is also strong momentum around open digital public infrastructure and open-source systems, enabling transparency, adaptability, and local ownership particularly important for smaller states. Across these examples, the key insight is that effective AI governance is not just policy it must be embedded in systems, led by communities, and designed for sovereignty and sustainability.