AI governance must move from principles to practical action, UN dialogue hears

Bridging the global AI divide will require much more than expanding access to AI tools, participants heard during a thematic session of the United Nations Global Dialogue on AI Governance. Speakers argued that countries need digital infrastructure, reliable electricity, skilled workforces, trusted institutions and governance capacity if they are to shape AI on their own terms rather than simply consume technologies developed elsewhere.

Throughout the discussion, governments, UN agencies, academics and industry representatives stressed that the next phase of AI governance should focus on implementation. They called for stronger international cooperation, investment in local capabilities and practical measures to ensure AI contributes to sustainable development instead of reinforcing existing inequalities.

Capacity building means creating AI, not just using it

Opening the session, Robert Opp of the UN Development Programme (UNDP) argued that the world is moving from a digital divide to an AI divide, one shaped not only by access to technology but also by countries’ ability to adopt, govern and develop AI responsibly.

Loretta Hieber Girardet of the UN Office for Disaster Risk Reduction (UNDRR) added that governments need trusted institutions, robust data systems and technical expertise if AI is to improve disaster resilience and public services.

The session’s co-chairs, Rashid Khan, Co-Founder of Yellow.ai, and Mark Alexandre Doumba, Gabon’s Minister of Digital Economy and Innovation, reinforced that message by arguing that AI governance should now move beyond high-level principles towards practical action. Khan said the challenge is no longer agreeing that AI should be inclusive and trustworthy, but creating the standards, infrastructure and skills needed to make those principles meaningful.

Doumba argued that developing countries should not try to replicate the resource-intensive path taken by major AI powers. Instead, they should build AI ecosystems suited to their own economies, languages and cultural contexts.

‘We should not measure success by who builds the biggest models,’ he suggested, but by whether AI creates jobs, improves public services and supports local innovation.

Several participants also stressed that capacity development must extend far beyond basic AI literacy. Shikoh Gitau argued that countries should become creators of AI rather than passive users, describing the goal as building AI ‘for us, by us’. That requires investment not only in technical skills, but also in research, standards, financing and local entrepreneurial ecosystems.

Government representatives echoed that assessment. Speakers from South Africa, Bangladesh, Nepal, Oman and Ethiopia all identified electricity, connectivity, computing power, public-sector capacity and access to quality data as essential foundations for meaningful AI participation.

Environmental sustainability moves to the centre of AI governance

One of the strongest themes throughout the discussion was that environmental sustainability should no longer be treated as a secondary issue in AI governance.

The UN Environment Programme (UNEP) argued that AI depends on extensive use of electricity, water, minerals and manufacturing while also generating growing volumes of electronic waste. Because these impacts extend across entire supply chains, speakers said governance should address AI’s full environmental lifecycle rather than focusing solely on the operation of AI models.

Participants also highlighted questions of environmental justice. Several speakers warned that many of the environmental costs associated with AI infrastructure, including mining, water consumption and waste, are disproportionately borne by communities in developing countries that receive relatively few of AI’s economic benefits.

Rather than assuming AI will automatically solve environmental challenges, panellists called for internationally comparable methods to measure AI’s environmental footprint, greater transparency from technology companies and stronger accountability across supply chains.

The discussion reflected a broader shift in international AI policy debates, with environmental sustainability increasingly treated as a core governance issue alongside safety, human rights and economic development.

Local languages and cultures must shape AI development

Another recurring message was that AI will only become genuinely global if it better reflects the world’s linguistic and cultural diversity.

Estonian President Alar Karis described how Estonia has invested heavily to ensure that AI systems can operate effectively in the Estonian language, despite the country’s relatively small population. Alongside partnerships with companies such as OpenAI and Google, Estonia has focused on training teachers, integrating AI into education and ensuring that modern Estonian-language content remains available for future AI systems.

Other speakers argued that similar efforts are needed worldwide. They noted that current AI models overwhelmingly favour dominant languages, leaving thousands of languages and many indigenous knowledge systems largely excluded from the AI ecosystem.

Several participants warned that countries lacking local datasets, evaluation benchmarks and language resources risk becoming dependent on technologies designed for entirely different cultural contexts.

The discussion also highlighted the importance of standards and international cooperation. UNESCO presented its ongoing work to implement its Recommendation on the Ethics of Artificial Intelligence through large-scale training programmes, language-diversity initiatives and AI competency frameworks for teachers and public officials.

Meanwhile, standards experts argued that participation in international standard-setting should itself be viewed as a form of capacity development, enabling developing countries to help shape the technical foundations of future AI systems.

Trust, children’s rights and implementation now take priority

Beyond infrastructure and capacity, speakers repeatedly argued that trust will determine whether AI delivers a broad public benefit. Participants emphasised that trustworthy AI requires transparent governance, accountable institutions and meaningful public oversight rather than technical performance alone.

Children’s rights received particular attention during the session. UNICEF warned that children are adopting AI technologies faster than adults are learning to regulate them, creating new risks around privacy, safety and development. Representatives called for child-centred benchmarks, stronger safeguards for children’s data and mandatory child-rights impact assessments for AI systems deployed in education, healthcare and other public services.

Several speakers also argued that governance should focus more on AI deployment than on frontier model development alone, ensuring that systems remain accountable throughout their lifecycle and can be adapted to local social and institutional realities.

Closing the session, Khan and Doumba returned to the discussion’s central message: that AI governance should ultimately be judged by practical outcomes rather than technological competition. Countries need the capability to shape AI according to their own priorities, they said, while international cooperation should ensure that no society is left behind.

Participants were encouraged to leave Geneva not simply with new principles, but with concrete commitments on financing, infrastructure, skills and cooperation that can be reviewed when the Global Dialogue reconvenes in 2027.

Track all key moments from the Global Dialogue on AI Governance inaugural meeting on our dedicated page.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!