Bankable Wisdom

Success would mean that the communities most affected by ungoverned AI are the ones whose experiences shape its governance, not simply referenced as statistics in reports written by others. For organizations like ours working in African EdTech, success looks like three concrete outcomes. First, binding commitments that AI systems deployed in education, employment screening, and digital finance must be tested for bias against populations with limited data footprints before deployment. Millions of young Africans are invisible to AI systems because they lack formal credit histories, standardized academic records, or English-language digital presence. This invisibility is not a neutral gap. It actively reproduces poverty. Second, enforceable transparency requirements so that learners, workers, and communities know when an AI system is making or influencing a decision that affects them, whether that is a loan application, a job screening, or a content recommendation algorithm shaping what skills they believe are worth learning. Third, a funded follow-up mechanism. The credibility of this Dialogue depends entirely on whether it produces action, not more consultation. Governments must leave Geneva in July with specific, time-bound commitments that are publicly tracked. One meeting without accountability structures is not governance. It is theater. The communities we work with have been promised inclusion in global processes before. What they need is not a seat at a table where decisions have already been made. They need to be in the room before the agenda is set.

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

Bankable Wisdom is an African EdTech platform training young people in income-generating digital skills across 30 African countries. Our learners are predominantly youth who have never been meaningfully represented in the datasets that power global AI systems. We selected AI capacity-building because the gap between communities that can use AI as a tool and communities that are simply subject to AI decisions is growing faster than any governance framework is closing it. For our learners in Lagos, Kano, and across West Africa, AI literacy is not optional. It is the difference between participating in the digital economy and being excluded from it entirely. We selected social and economic implications because AI is already reshaping the labor market our graduates are entering. Automated hiring tools, algorithmic content moderation, and AI-generated misinformation about financial opportunities are concrete daily realities, not theoretical risks. We selected human rights protection because our learners frequently encounter digital systems that flag them as high-risk, untrustworthy, or low-value on the basis of where they live, what language they speak, or what phone they use. This is a rights issue, not only a technical one. We selected transparency and accountability because when an AI system makes a wrong decision about one of our learners, there is currently no mechanism for them to know it happened, challenge it, or seek redress. Accountability without visibility is impossible.

The form does not adequately name the problem of AI systems trained on data that excludes the Global South by design. The majority of large language models, credit scoring tools, hiring algorithms, and content recommendation systems are trained predominantly on English-language, Western-context data. This is not a minor technical limitation. It means that AI systems routinely produce outputs that are irrelevant, incorrect, or harmful when applied to African contexts, languages, and realities. A second missing issue is the weaponization of AI-generated disinformation against economically vulnerable communities. In Nigeria and across West Africa, AI-generated fake testimonials, deepfake endorsements of fraudulent investment schemes, and algorithmically amplified scam content are actively destroying trust in legitimate digital education and financial services. This disinformation does not only harm individuals. It erodes the entire ecosystem of trust that platforms like ours depend on to operate. Third, the form does not address the question of who owns the economic value generated when AI systems are trained on community-generated content, local language data, or behavioral data harvested from users in developing countries. Communities whose data trains these systems receive none of the value created. This is an equity and sovereignty issue that governance frameworks have not yet named clearly.

In Nigeria's EdTech and digital skills sector, the absence of AI governance creates three overlapping harms. First, learners are subjected to AI-driven hiring tools and freelance platform algorithms that systematically deprioritize profiles from Nigerian IP addresses, regardless of skill level. Our graduates apply for remote work opportunities on global platforms and are filtered out before a human reviewer sees their application. There is no transparency about why, no appeal mechanism, and no accountability for the platform deploying the filter. Second, AI-generated financial scam content floods the same digital channels we use to reach learners, making it structurally harder for legitimate income-skills education to be seen and trusted. The asymmetry is striking: bad actors use AI to scale harmful content rapidly and cheaply, while legitimate educators bear the reputational cost. Third, African governments lack the technical capacity to evaluate AI systems being sold to them by global vendors for use in education, border control, and social services. In the absence of governance standards, procurement decisions are made without impact assessments, and communities have no recourse when outcomes are harmful. The opportunity is equally real. A properly governed AI ecosystem in Africa could unlock personalized learning at scale, connect youth with global employment without geographic discrimination, and formalize the informal economy. The governance gap is not only a protection problem. It is an economic one.

The AI Dialogue has an opportunity to do something no bilateral agreement or national regulation can accomplish alone: establish a shared minimum floor of protection that applies regardless of where an AI system is built or where it is deployed. For countries in the Global South, this matters enormously. When a hiring algorithm built in San Francisco or London is used to screen a candidate in Lagos or Nairobi, which jurisdiction's rules apply? Currently, the answer is effectively none. The developer faces no accountability in the market where the harm occurs, and the government in that market lacks the leverage to impose standards on a foreign technology company. The Dialogue should work toward recognition that AI systems, like environmental pollutants, can cause harm across borders, and that the entity deploying harm bears responsibility regardless of where that harm is experienced. This principle already exists in international environmental law. It needs to be established in AI governance before the harm compounds further.

The African Union's Continental AI Strategy and the AU's Digital Transformation Strategy 2020 to 2030 both articulate governance principles rooted in African contexts and priorities. The Dialogue should explicitly connect with these frameworks rather than positioning itself as the primary architecture that regional bodies must align to. UNDP's Timbuktoo initiative, which supports EdTech startups across Africa including Bankable Wisdom, represents a model of technology governance that is community-grounded, outcome-focused, and built with rather than for affected populations. This model should inform how the Dialogue approaches capacity-building, not as a transfer of technical knowledge from North to South, but as a recognition that communities already possess the contextual expertise necessary to govern technology in their own realities. The Dialogue should also connect with the work of civil society coalitions that have been doing AI rights and accountability advocacy for years before global institutions paid attention. Their analysis should be credited and built upon, not consulted and then set aside.

The format of the Dialogue must match its stated commitment to inclusion. A meeting held in Geneva in July, accessible primarily to those with institutional affiliations, travel budgets, and English fluency, will reproduce the exact exclusion it claims to address. Concrete recommendations: sessions must be available in real time in at least French, Arabic, Swahili, Portuguese, and Hausa, which together cover the majority of the African digital population. Asynchronous participation pathways must be funded and promoted with the same energy as in-person attendance. Community organizations without UN accreditation must have a formal, non-decorative pathway to submit inputs that are read and cited in official outputs. The private sector's participation should be structured with conflict of interest disclosure requirements. A company building AI systems that will be governed by the Dialogue's outputs should not have the same participatory weight as a civil society organization representing communities harmed by ungoverned AI.

Young people in informal economies across Africa, Southeast Asia, and Latin America who use AI tools daily but have never been asked their opinion about how those tools should be governed. Gig workers whose livelihoods are shaped by algorithmic management systems they cannot see or challenge. Women entrepreneurs in digital markets who face compounded discrimination from both gender bias in training data and the economic precarity that limits their ability to participate in formal advocacy processes. Indigenous and minority language communities whose languages are absent from the systems making decisions about them. These communities are not hard to reach. They are simply not being prioritized. The Dialogue should publish a transparency report after it concludes, showing who contributed, from which regions, in which languages, and through which mechanisms. Accountability for inclusion must be measurable, not aspirational.

The most valuable engagement format would be structured testimonial sessions where communities most affected by ungoverned AI present specific documented cases directly to government delegates, without interpretation by intermediaries, advocacy organizations, or UN staff. Governments need to hear from a young woman in Abuja who was rejected by an AI hiring tool she never knew existed. They need to hear from a learner whose digital skills certificate was flagged as fraudulent by an automated verification system that had never been trained on African educational institutions. These are not hypothetical harms. They are occurring now, and the people experiencing them are rarely in rooms where policy is made. A pre-Dialogue open submission period with public access to all submissions, not only a curated selection, would also significantly increase legitimacy and accountability. If the process claims to be consultative, the full record of consultation should be publicly available.

The NITDA-Coursera partnership in Nigeria, which has provided government-structured digital skills training to over one million Nigerians, demonstrates that government mandates can dramatically accelerate AI and digital capacity-building when they are paired with accessible, locally relevant content. The model is worth scaling, with the addition of outcome tracking and accountability for actual employment and income impact, not only course completion numbers. Brazil's use of algorithmic impact assessments before deploying AI in public sector decisions offers a replicable governance model that smaller governments can adapt without requiring extensive technical infrastructure. The key principle, that systems must be evaluated for discriminatory impact before deployment, not after harm has occurred, is straightforward and should become a global minimum standard. Bankable Wisdom's own model of community-grounded, income-outcome-focused digital education offers a small but concrete example of what human-centered AI capacity-building looks like in practice: start with what the learner needs to earn, not what the technology can do, and build from there.