Governance, with Grit

The Dialogue succeeds if it shifts the centre of gravity. AI governance conversations remain dominated by frontier model risks and the priorities of a handful of technology-exporting nations. For most of the world, AI means something different: food security, drug safety, maternal health, access to education. The Dialogue should reflect that. In Malawi, farmers send WhatsApp voice notes in Chichewa to an AI chatbot that identifies crop diseases from photographs. In Senegal, voice-based AI delivers agronomic advice in local languages where one extension agent serves 15,000 producers. In Nigeria, mobile authentication services and AI-powered scanners have helped reduce counterfeit antimalarial drugs from nearly 20% to under 4%. These are governance successes - built on local knowledge, shaped by local constraints, running on existing infrastructure. A governance framework designed around billion-parameter models and data centre procurement misses this. The Dialogue should produce outcomes that treat these applications as reference points, not footnotes. Three markers denote success: capacity building, open source, and interoperability. 1. Capacity-building under cluster centres on enabling countries to govern AI on their own terms - not importing frameworks wholesale. 2. Open-source and open-data commitments under draw on equitable licensing precedents, where legal frameworks have enabled life-saving technologies to reach the populations that need them - as happened when patent renegotiation cut the cost of HIV medication by over 90% in South Africa. 3. Interoperability under accounts for the infrastructure realities of low-connectivity, voice-first, multilingual environments where AI already delivers measurable impact. Governance shaped by those who govern themselves produces better outcomes than governance exported from elsewhere.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Interoperability of governance approaches
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

Four clusters demand attention, in sequence, because each builds on the previous one and together they interconnect. 1. The social, economic, cultural, linguistic and technical implications of AI. AI training data reflects the demographics of its origins: predominantly English-language, predominantly male, predominantly from high-income countries. This shapes what AI systems produce, whom they serve, and whose knowledge they encode. The implications extend beyond data. The physical infrastructure of AI - from cobalt and lithium extraction in the Democratic Republic of Congo to rare earth processing concentrated in China to the helium required for semiconductor fabrication, a third of which disappeared from global supply when strikes hit Qatar's Ras Laffan facility in February - maps directly onto the UN Sustainable Development Goals. Minerals extracted from lower-income countries power hardware manufactured and sold back at margins those countries cannot capture. Ignoring this lifecycle only governs the surface. 2. Interoperability of governance approaches. AI operates across jurisdictions. Its impacts intersect liability, intellectual property, environmental, competition, and human rights law simultaneously. When private companies determine the behavioural parameters of systems used by billions, those decisions carry regulatory force without regulatory accountability. Interoperability requires more than mutual recognition of standards; it requires shared understanding of the non-negotiable areas governance must address. 3. Protection and promotion of human rights. The first two clusters lead here. Training data that excludes populations produces systems that exclude populations. Supply chains that rely on forced labour and environmental destruction in extracting countries undermine the rights the Dialogue seeks to protect. 4. Transparency, accountability and human oversight. Without mechanisms to audit what systems do, whom they affect, and how decisions propagate across borders, the preceding commitments remain aspirational. The four clusters interconnect. Addressing them in isolation reproduces the fragmentation this Dialogue exists to overcome.

The listed themes address what AI does and how to govern it. They do not address who gets to decide. The Dialogue should name a cross-cutting principle that underpins all four clusters: the right of peoples and nations to digital self-determination in AI. This extends existing international law. Article 1 of both the International Covenant on Civil and Political Rights and the International Covenant on Economic, Social and Cultural Rights recognises the right of all peoples to freely determine their political status and freely pursue their economic, social, and cultural development. AI systems now shape all three. Training data determines whose language, knowledge, and culture AI encodes. Infrastructure ownership determines who controls the computational layer on which public services increasingly depend. Behavioural parameters set by private companies determine the information billions receive. Without extending self-determination into these domains, governance risks becoming something done to countries rather than by them. The evidence that this works already exists. In 2007, Kenya's M-Pesa demonstrated that technology designed around local constraints - SMS-based, no bank account required, agent networks built on existing social infrastructure - could leapfrog systems designed elsewhere. Kenya's Central Bank enabled this through a regulatory posture now studied internationally: permitting innovation within a framework rather than importing one. From M-Pesa, the pattern continues. Nigeria's RxAll uses AI-powered scanners to verify drug authenticity at point of purchase. Malawi's UlangiziAI delivers agricultural advice through WhatsApp voice notes in Chichewa. These are not adaptations of imported systems. They originate from the communities they serve. The Dialogue should recognise this pattern as a governance principle. Countries that determine how AI operates within their borders - shaping it around their languages, infrastructure, and priorities - produce systems that work. Digital self-determination belongs in the architecture of this Dialogue.

I am a social researcher. Listening sits at the centre of my practice - across policy, regulatory, and organisational contexts. I am also neurodivergent, which extends the listening further: patience that outlasts discomfort, pattern recognition across domains others compartmentalise, and a tendency to take people at their word. My region - Western Europe - already has functioning food security, drug regulation, and education infrastructure. AI governance conversations here focus on productivity, competition, and frontier model risk. For much of the world, AI addresses needs that remain unmet - and governance gaps determine whether it reaches the people who need it. The challenge: governance frameworks exported from Western Europe and North America encode priorities, infrastructure assumptions, and risk appetites based on Western norms. They fit poorly where the norms and baselines differ. The opportunity: communities already building AI around local languages, local infrastructure, and local needs - as the examples in my previous responses demonstrate - produce governance evidence that works. That evidence exists. It needs collating, It needs a platform.