Blockchain & Climate Institute (BCI)

Success for the first Global Dialogue on AI Governance cannot be measured by the quality of its communiqué. It must be measured by what changes after the room empties. Four outcomes would mark this Dialogue as genuinely consequential. 1. A Global AI Governance Registry: the international community currently lacks a single, authoritative map of what each country has actually built — which frameworks exist, where the gaps are, and which approaches are compatible across borders. A living, publicly accessible registry would transform visibility into accountability and make interoperability a practical goal rather than an aspirational one. 2. A binding AI Capacity Compact: advanced economies, technology companies, and multilateral institutions must make specific, measurable commitments to transfer compute access, technical talent, and data infrastructure to developing nations. Not as development aid — as a strategic imperative. An AI divide today is a governance crisis tomorrow. 3. A Cross-border AI Incident Disclosure Protocol: we have voluntary, standardized mechanisms for reporting harm in aviation, pharmaceuticals, and cybersecurity. AI governance has no equivalent. A voluntary but structured multilateral protocol — covering financial services, healthcare, and public administration — would create the evidence base the international community needs to govern effectively rather than reactively. 4. Elevation of the Global South as a governance source, not just a case study: the most durable AI governance frameworks will be those tested at scale in resource-constrained, high-stakes environments. Africa's mobile financial infrastructure was built before global regulatory consensus existed — and the world has been learning from it ever since. This Dialogue should institutionalize that dynamic through a Global South Lighthouse Initiative that amplifies emerging-market governance innovations globally. The Scientific Panel's inaugural report must drive each of these, not decorate the agenda. Delegates should leave Geneva with owners, timelines, and accountability mechanisms attached to every commitment made. That is what success looks like.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Transparency, accountability, and human oversight
• Interoperability of governance approaches

The selection reflects a single organizing principle: climate action and AI governance are no longer parallel conversations. They are the same conversation, and the stakes of getting it wrong are planetary. 1. Safe, secure and trustworthy AI sits at the foundation of everything BCI does. Climate data infrastructure - carbon credit verification, ESG disclosure, adaptation risk modelling - is only as credible as the AI systems processing it. Markets and governments are already making trillion-dollar decisions on AI-assisted climate outputs. Trust is not a feature; it is the entire value proposition. 2. Transparency, accountability, and human oversight follow directly. BCI has observed, across multiple jurisdictions, how AI is being embedded into climate finance allocation and carbon market governance with insufficient auditability. When an algorithm influences which communities receive adaptation funding or which emissions reductions are certified, that algorithm must be explainable, contestable, and supervised by accountable humans. Opacity in climate 3. AI is not just a governance failure - it is an equity failure. Social, economic, and ethical implications matter profoundly to BCI because climate vulnerability and AI exclusion frequently overlap. The communities most exposed to climate risk are often the least represented in training data, the least consulted in system design, and the least protected when AI-driven decisions go wrong. Our field work across Africa and the Pacific makes this reality impossible to ignore. 4. Interoperability of governance approaches is where BCI's blockchain and climate policy expertise converges most directly with AI governance. Carbon markets, climate disclosure regimes, and adaptation finance flows are inherently transboundary. If AI governance frameworks remain nationally siloed, climate actors will face a fragmented regulatory landscape that creates arbitrage, slows ambition, and erodes the coherence that effective climate finance demands.

Yes. Three issues deserve explicit recognition. AI and Climate Infrastructure as Critical Systems. None of the seven themes specifically addresses the growing role of AI in managing physical climate infrastructure - flood early warning systems, drought prediction models, grid balancing for renewable energy, and sea-level monitoring networks. These are not consumer applications. They are critical systems on which lives and livelihoods depend. When they fail, communities - overwhelmingly in the Global South - bear the consequences. The Dialogue should establish a dedicated framework for AI deployed in climate-critical infrastructure, with commensurate standards for resilience, redundancy, and accountability. The Carbon Cost of AI Itself. The environmental footprint of large-scale AI - data centre energy consumption, water usage for cooling, hardware supply chains - is growing faster than the governance conversation acknowledges. BCI is concerned that as the international community races to deploy AI for climate solutions, it is simultaneously generating a new and largely unaccounted emissions category. This is not an argument against AI; it is an argument for honest accounting. The Dialogue should initiate a methodology for AI lifecycle emissions disclosure, consistent with existing climate reporting frameworks. Algorithmic Capture of Climate Finance. As AI becomes embedded in credit scoring, risk modelling, and investment screening within climate finance institutions, there is a structural risk that historical patterns of exclusion are reproduced and accelerated at scale. Communities and projects that have been chronically underfinanced may find themselves systematically deprioritized by AI systems trained on data that reflects past inequity rather than future potential. This is a form of governance failure with no clear home in the current thematic architecture. It sits at the intersection of human rights, economic implications, and accountability - but belongs to none of them fully.

The governance gaps are not theoretical for Kenya and the African continent. They are showing up in real systems, real markets, and real communities right now. The most significant challenge is ungoverned AI in carbon and climate finance. Africa hosts a disproportionate share of the world's voluntary carbon market projects — forestry, soil carbon, clean cookstoves, renewable energy. AI is increasingly used to verify, monitor, and price these credits. Yet there is no agreed standard for how these systems are validated, audited, or contested. Kenyan smallholder farmers participating in soil carbon programmes have no visibility into how AI-generated measurements affect their income. That is a governance gap with immediate economic consequences. Data exclusion is compounding climate vulnerability. African weather stations are sparse. Historical climate datasets are thin. AI models trained predominantly on data from the Global North produce adaptation recommendations that are poorly calibrated for East African agroecological zones, rainfall variability, and informal settlement patterns. The result is that AI-assisted climate planning tools systematically underserve the populations most exposed to climate risk — precisely because the governance frameworks that should mandate representative data did not exist when these systems were built. The opportunity is equally significant. Kenya's technology infrastructure — mobile-first financial systems, a growing AI and blockchain developer community, and hard-won experience governing digital innovation at scale — positions the country as a credible exporter of governance models, not merely an importer of technology. The M-Pesa regulatory experience, the Sandbox frameworks of the Central Bank of Kenya represent practical governance innovations the international community can learn from. Africa does not need to wait for Geneva to lead. But Geneva must create the architecture that makes African leadership count globally. That is the gap this Dialogue must close.

The AI Dialogue's most consequential role is one that no existing institution currently performs: serving as the permanent connective tissue between AI governance and the full breadth of international policy agendas — climate, trade, finance, health, and human rights — rather than treating AI as a standalone technical domain. Three specific roles would make the Dialogue genuinely indispensable. First, it must function as a norm convergence platform, not a norm production factory. The international community does not lack AI principles. It lacks mechanisms to reconcile competing national and regional frameworks into workable interoperability agreements. The Dialogue is uniquely positioned — through its multi-stakeholder architecture and UN convening authority — to facilitate the kind of behind-the-scenes alignment that turns parallel frameworks into compatible ones. For Africa, where multiple overlapping governance initiatives are simultaneously underway, this convergence function is urgent. Second, it must institutionalize the Global South's role as a governance co-creator. This means structured mechanisms — not token representation — for emerging economies to shape foundational standards before they are exported as fait accompli. My view is that African nations, which are simultaneously among the most AI-vulnerable and the most governance-innovative, must have formal co-authorship over the frameworks that will govern AI in their contexts. The Dialogue's multi-stakeholder model creates the architecture for this. The question is whether the political will exists to activate it. Third, it must create accountability infrastructure between sessions. The most persistent failure of international governance dialogues is that commitments made in plenary evaporate before implementation. The Dialogue should establish a between-session accountability mechanism — linked to the Scientific Panel's ongoing work — that tracks progress, names gaps, and maintains pressure on parties between annual convening cycles. Done well, the Dialogue becomes the institution the world needed before AI governance fragmented. That window remains open, but not indefinitely.

Several consequential initiatives are already operating in this space. The Dialogue's value lies not in duplicating them but in doing what none of them can do alone — connecting them into a coherent international architecture. Initiatives the Dialogue should build upon: 1. The OECD AI Policy Observatory has produced the most comprehensive comparative mapping of national AI frameworks currently available. Rather than replicating this work, the Dialogue should formally integrate it as the foundation of the proposed Global AI Governance Registry, adding the multilateral legitimacy and developing-country coverage the OECD's membership structure cannot provide. 2. The African Union's Continental AI Strategy and the Smart Africa Alliance represent serious, regionally-owned governance work that the international community consistently overlooks. The Dialogue should treat these not as regional inputs to a global conversation but as co-equal frameworks deserving mutual recognition and integration. 3. The UNFCCC's work on digital innovation for climate action — including AI applications in National Adaptation Plans and carbon market integrity — intersects directly with BCI's mandate. The Dialogue should establish a formal liaison mechanism with the UNFCCC Secretariat to ensure AI governance standards developed in Geneva are compatible with climate reporting obligations under the Paris Agreement. 4. The ITU's AI for Good platform and UNESCO's Recommendation on the Ethics of AI both carry significant implementation infrastructure. The Dialogue should leverage rather than compete with these, using them as delivery mechanisms for capacity-building commitments made under the proposed AI Capacity Compact. The Dialogue's specific added value is its General Assembly mandate — the political authority to make decisions that bind member states and create accountability across all of these parallel tracks. Every existing initiative operates within a bounded constituency. The Dialogue alone has the convening authority to make the connections between them compulsory rather than voluntary. That authority should be used deliberately and urgently.

The Dialogue's format will determine whether its outputs are owned by governments alone or by the full range of actors whose decisions actually shape how AI develops and governs itself. Structure is not a procedural question — it is a power question. On stakeholder contribution: 1. Member States must move beyond position-stating into genuine negotiation. The Dialogue should introduce structured bilateral and small-group working sessions between plenary meetings — modelled on the UNFCCC's contact group format — where real convergence on text and commitments can happen away from formal proceedings. 2. Private sector actors — particularly frontier AI developers — should be required to submit transparency disclosures as a condition of participation, not as a voluntary gesture. Their technical knowledge is indispensable; their accountability to the process must be commensurate with their influence over it. 3. Civil society and affected communities require more than speaking slots. The Dialogue should establish a formal Civil Society Advisory Panel with defined rights to propose agenda items, respond to draft outputs, and access working documents on equal terms with member state delegations. This specifically means climate-vulnerable communities having structured input into AI governance decisions that directly affect adaptation planning and climate finance access. 4. Technical and scientific community input should flow through the Scientific Panel with a direct and transparent line to negotiated outcomes — not filtered through diplomatic summaries that dilute technical precision. On format and structure: The Dialogue should adopt a three-tier model: an annual high-level plenary for political commitment, quarterly thematic working groups for substantive technical progress, and a continuous digital participation platform enabling asynchronous contribution from stakeholders in all time zones and with limited travel budgets. Sessions should alternate between Geneva and regional host cities — beginning with an African host — to signal genuine decentralization of authority rather than its performance. Inclusion must be structural, not ceremonial.

The gap between who governs AI globally and who bears its consequences is the most fundamental legitimacy problem the Dialogue faces. Closing it requires structural solutions, not better optics. The most underrepresented voices: 1. Smallholder farmers and rural communities in the Global South are among the most directly affected by AI governance decisions — through climate adaptation tools, agricultural insurance algorithms, and financial inclusion systems — yet have no pathway into international governance processes. Their exclusion is not incidental. It is structural. These communities lack the institutional affiliation, travel budgets, and formal credentials that current participation models require. 2. Indigenous communities hold irreplaceable knowledge systems relevant to climate adaptation, biodiversity monitoring, and land governance — all domains where AI is being rapidly deployed. Yet AI systems are being trained on data extracted from these communities without consent, and governance frameworks are being designed without their input. This is simultaneously an ethical failure and a technical one — the resulting systems are worse for the exclusion. 3. Informal economy workers — who represent the majority of employment across Africa and much of Asia — face AI-driven disruption in labour markets, credit access, and public service delivery with no representation in the governance conversations shaping those systems. Young people in low-income countries will live longest with the consequences of decisions being made now, yet youth engagement mechanisms in international governance remain largely tokenistic. How to include them: The Dialogue should establish a Distributed Voices Programme — a funded mechanism enabling civil society organisations rooted in underrepresented communities to participate substantively, not symbolically. This means travel funding, translation infrastructure, asynchronous contribution pathways, and crucially, formal rights to influence outputs rather than merely inform them. Beyond access, the Dialogue must address authority. Representation without decision-making power reproduces the exclusion it claims to remedy. Presence is not participation. Participation is not power. The Dialogue must deliver all three.

The formats that will define this Dialogue are not the ones that fill the programme — they are the ones that change what participants do after they leave the room. That requires deliberate design, not convention. Structured Adversarial Sessions. Rather than consecutive presentations of aligned positions, the Dialogue should introduce formally moderated adversarial exchanges — pairing a frontier AI developer with a climate-vulnerable community representative, or a Global North regulator with a Global South innovator — with a mandate to surface genuine disagreement rather than diplomatic consensus. Real tension, honestly held, produces better governance than managed agreement. Scenario Stress-Testing Workshops. Participants should be confronted with specific, plausible AI governance failures — a carbon market manipulated by an opaque verification algorithm, an early warning system that failed a flood-prone community due to biased training data — and required to work through response protocols in real time. Governance designed against concrete failure scenarios is more durable than governance designed around abstract principles. Citizen Deliberative Panels. Randomly selected members of affected communities — farmers, informal workers, young people from AI-vulnerable regions — should participate in structured deliberative sessions with formal rights to question delegations and respond to draft outputs. This is not symbolic inclusion. Deliberative democracy at this scale changes both the quality of outputs and their perceived legitimacy. Open Working Document Sessions. Draft texts and commitment frameworks should be developed in live, observable sessions — streamed publicly with structured digital commentary from registered stakeholders. Transparency in the drafting process creates accountability before documents are finalized, not after. Between-Session Digital Deliberation. A purpose-built multilingual platform should sustain structured dialogue between annual convenings — not social media, but facilitated deliberation with synthesis mechanisms that feed directly into formal proceedings. The best governance conversations do not happen in plenary. The Dialogue's formats should be designed around that truth.

The most instructive examples of effective AI governance share a common characteristic: they were built from operational reality, not regulatory theory. The following represent concrete models the Dialogue should study and scale. Kenya's Regulatory Sandbox Architecture. The Central Bank of Kenya's fintech sandbox framework demonstrated that governance and innovation are not opposites - they are partners when regulation is designed to learn alongside technology rather than chase it. This model, offers a replicable template for jurisdictions seeking to govern emerging technology without suppressing it. The EU AI Act's Risk-Based Tiering. Despite its imperfections, the Act established a globally significant precedent: that AI governance should be calibrated to consequence, not to technology category. High-risk applications face stringent requirements; low-risk applications face minimal friction. This proportionality principle is exportable and should inform the Dialogue's framework recommendations, particularly for climate-critical AI applications that currently lack equivalent classification. CGIAR's Open Climate Data Infrastructure. The Consultative Group on International Agricultural Research has built open, interoperable agricultural and climate datasets accessible to AI developers across the Global South. This represents the kind of open data commons the Dialogue should actively replicate and fund - reducing the data exclusion that currently produces AI systems poorly calibrated for African and Asian agroecological contexts. These are not aspirational. They are operational. The Dialogue should treat them as its starting point.