Ministry of Preschool and School Education

A successful first Global Dialogue on AI Governance should not be judged by how comprehensive its discussions are, but by whether it produces practical alignment, inclusive ownership, and credible next steps. First, success would mean achieving a "minimum viable consensus" rather than an overly ambitious universal framework. Given geopolitical differences, expecting full agreement is unrealistic. Instead, the dialogue should identify 4-5 non-negotiable commitments, such as human accountability in AI decision-making, transparency in high-risk systems, and safeguards against harm, that all participants can endorse and realistically implement. Second, the dialogue must demonstrate genuine inclusion beyond representation. It is not enough to have participants from the Global South; their perspectives should shape outcomes. For example, priorities like AI in public education, multilingual access, and digital equity should be reflected in final recommendations. This would ensure governance is not export-driven but co-created. Third, a defining outcome would be the creation of sector-specific pilot initiatives. Rather than broad promises, the dialogue should launch 2-3 collaborative pilots (e.g., AI in education, public health, or workforce training) involving multiple countries. These pilots would act as testing grounds for governance principles, making the dialogue action-oriented and evidence-based. Fourth, success would involve establishing a light but durable coordination mechanism, such as a rotating multi-stakeholder working group, to track progress, share best practices, and adapt guidelines over time. This avoids both fragmentation and bureaucratic stagnation. Finally, the dialogue should result in a shared narrative of responsible AI that is accessible to non-experts, particularly educators, students, and communities. Governance cannot remain confined to policy circles; it must translate into awareness and practice at the grassroots level. In essence, the dialogue would be successful if it shifts global AI governance from abstract debate to tested cooperation, grounded in realism, inclusivity, and continuous learning.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• AI capacity-building
• Transparency, accountability, and human oversight
• Protection and promotion of human rights

My priorities are shaped by my work at the intersection of education, language, and AI integration, particularly in diverse and multilingual learning environments. AI capacity-building is the most urgent priority because the gap between access to AI tools and the ability to use them meaningfully is widening. Educators, students, and institutions, especially in developing contexts, require structured training, digital literacy, and pedagogical support to engage with AI responsibly and effectively. The social, cultural, and linguistic implications of AI are equally critical. AI systems often reflect dominant languages and cultural norms, which risks marginalising learners from diverse linguistic and socio-cultural backgrounds. Ensuring inclusivity in AI design and deployment is essential, particularly in education systems where language and identity are deeply interconnected. Transparency, accountability, and human oversight are necessary to build trust in AI systems. In educational settings, for instance, decisions influenced by AI, such as assessment or feedback, must remain interpretable and subject to human judgment to avoid bias and over-reliance on automated outputs. Finally, the protection and promotion of human rights provides the ethical foundation for all AI governance efforts. Issues such as data privacy, equitable access, and non-discrimination must remain central, especially when AI is deployed in public systems like education. Together, these priorities emphasise not only technological advancement but also equitable, inclusive, and responsible AI adoption.

One key cross-cutting issue not fully captured is the impact of AI on education systems and learning ecosystems. While elements of this appear across themes, the scale and speed at which AI is reshaping teaching, assessment, and knowledge production require more focused global attention. Questions around academic integrity, teacher roles, and student dependency on AI tools are rapidly emerging. Another critical issue is AI-driven inequality within and across countries. Beyond access, there is a growing divide in terms of who can shape, localise, and govern AI systems. Without deliberate intervention, this could deepen existing socio-economic and digital inequalities. The professional development and well-being of educators and workers adapting to AI is also underexplored. As AI transforms workplaces, continuous upskilling and psychological readiness will be essential, particularly in sectors like education where human interaction remains central. Finally, context-sensitive governance is an emerging need. Global frameworks often struggle to translate into local realities. There is a need for adaptable governance models that respect cultural, linguistic, and institutional diversity while maintaining shared global standards. Addressing these cross-cutting issues will ensure that AI governance remains responsive, inclusive, and grounded in real-world challenges.

In my context, particularly within the education sector in emerging and multilingual settings, gaps in AI governance are already producing uneven outcomes that reflect both significant challenges and opportunities. One of the most pressing challenges is the disparity between rapid AI adoption and limited capacity-building, where educators and institutions often lack the training to critically and effectively integrate AI into teaching and learning. This is compounded by insufficient attention to linguistic and cultural diversity, as many AI tools are designed primarily for dominant languages, limiting their relevance and inclusivity in local classrooms. At the same time, weak frameworks around transparency and human oversight create risks of over-reliance on AI-generated feedback, raising concerns about bias, academic integrity, and the erosion of professional judgment. From a human rights perspective, issues such as data privacy, unequal access to digital infrastructure, and the potential marginalisation of vulnerable learners remain significant. However, these gaps also present clear opportunities: they create space for developing context-sensitive governance models that prioritise inclusive design, multilingual accessibility, and educator empowerment. With targeted investment in capacity-building and clearer accountability mechanisms, AI can enhance personalised learning, expand access to quality education, and support teachers rather than replace them. Overall, the current moment represents a critical window to shape AI governance in a way that aligns technological advancement with equity, local relevance, and long-term societal benefit.

The AI Dialogue can play a pivotal role by acting as a bridge between global ambition and local implementation. Rather than becoming another high-level discussion platform, it can enable practical cooperation by aligning diverse stakeholders, governments, educators, technologists, and civil society around shared, actionable priorities. One of its most valuable contributions would be to foster trust-based collaboration, particularly between technologically advanced countries and those still developing AI ecosystems, ensuring that governance is co-created rather than imposed. The Dialogue can also serve as a space to translate global principles into adaptable frameworks, supporting countries in contextualising AI governance according to their social, cultural, and institutional realities. Additionally, it can promote knowledge exchange and capacity-sharing, where best practices, case studies, and policy experiments, especially from sectors like education, are openly shared and refined. By encouraging multi-stakeholder pilot initiatives, the Dialogue can move cooperation beyond statements toward real-world testing of governance models. Ultimately, its role should be to institutionalise continuous engagement, ensuring that AI governance evolves collaboratively, remains inclusive, and responds effectively to emerging challenges rather than becoming fragmented or reactive.

Several existing initiatives and partnerships already provide a strong foundation for advancing international AI governance, though their impact varies across regions. Multilateral efforts such as UNESCO's global recommendations on AI ethics and the OECD's AI principles have contributed to establishing shared normative frameworks, while initiatives like the Global Partnership on AI (GPAI) promote collaboration between governments and experts on responsible AI development. Regional frameworks, including the European Union's AI Act, offer more concrete regulatory approaches that can inform global discussions, even if they are not universally applicable. In parallel, there are growing cross-sector collaborations, particularly in education and digital capacity-building, where institutions, NGOs, and technology providers work together to improve AI literacy and equitable access. However, many of these mechanisms remain fragmented or unevenly accessible, especially for developing countries. This highlights the need for stronger coordination and more inclusive participation. Building on these existing efforts, future cooperation should focus on connecting global principles with local implementation, supporting capacity-building, and ensuring that diverse perspectives, particularly from underrepresented regions, actively shape the evolution of AI governance.

Different stakeholders can contribute meaningfully if the AI Dialogue is designed as a co-creation space rather than a consultation forum. Governments can anchor discussions in policy feasibility, while educators, civil society, and community practitioners can ensure that outcomes reflect real-world needs and constraints. The private sector and technical experts can provide insights into system design and deployment, but their contributions should be balanced with ethical and societal perspectives. Structurally, the Dialogue should combine focused thematic tracks with small, mixed-stakeholder working groups tasked with producing concise, actionable outputs (e.g., policy briefs or pilot proposals). Instead of one-way panels, each session should require collaborative input, ensuring that all participants actively shape outcomes. Allowing stakeholders to submit short evidence notes or case examples in advance would further enrich discussions and make participation more substantive.

Voices that remain consistently underrepresented include frontline educators, local policymakers, youth, and communities from linguistically diverse and low-resource settings. These groups are often the end-users of AI systems, but rarely influence how such systems are governed. Their exclusion risks creating governance models that are technically sound but socially disconnected. To address this, inclusion should be operationalised through targeted outreach and structural integration, not just invitations. This could include dedicated quotas for underrepresented groups in core sessions, multilingual participation formats, and partnerships with local institutions to identify contributors with contextual expertise. Importantly, these stakeholders should be involved in decision-shaping roles, such as rapporteurs or co-chairs of working groups, so their perspectives directly influence outcomes. Sustained engagement mechanisms, such as follow-up networks or advisory pools, would ensure their contributions continue beyond the Dialogue itself.

The AI Dialogue would benefit from formats that prioritise interaction, problem-solving, and shared accountability. One effective approach is "live governance labs," where participants work in small groups to design responses to realistic AI challenges, producing tangible outputs within the session. Another is reverse panels, where policymakers and technical experts primarily listen while educators, youth, and community representatives present lived experiences and practical concerns. Rotating roundtables can also encourage participants to engage with diverse perspectives rather than remaining in fixed groups. Additionally, incorporating real-time collaborative tools, such as shared digital boards for drafting recommendations, can make discussions more transparent and participatory. Short, time-bound "solution sprints" focused on specific issues (e.g., AI in education or data ethics) can further ensure that engagement leads to concrete ideas. These formats would transform the Dialogue from a series of discussions into a space for joint problem-solving and actionable outcomes.

Effective AI governance is already emerging through a combination of policy frameworks, sector-specific practices, and collaborative platforms that balance innovation with accountability. One notable example is the adoption of risk-based regulatory approaches, where AI systems are classified according to their potential impact, allowing stricter oversight for high-risk applications such as education, healthcare, and public decision-making. This model supports proportional governance while avoiding unnecessary barriers to innovation. In practice, human-in-the-loop systems have proven effective, particularly in education, where AI-assisted assessment tools are used alongside teacher judgment to ensure fairness, contextual understanding, and accountability. Another promising approach is the development of AI ethics guidelines integrated into institutional policies, where organisations embed principles such as transparency, fairness, and data protection into everyday workflows rather than treating them as abstract commitments. Platforms that promote open collaboration and knowledge-sharing, including cross-sector partnerships between governments, educational institutions, and technology providers, also play a critical role by enabling the exchange of best practices and locally adaptable solutions. Additionally, capacity-building initiatives, such as teacher training programs on AI literacy and responsible use, demonstrate how governance can be operationalised at the ground level. Together, these examples highlight that effective AI governance is not defined by a single model, but by a combination of adaptable policies, participatory practices, and continuous learning mechanisms that ensure AI systems remain aligned with human values and societal needs.