A landmark index measuring responsible AI across 138 countries has found that the people with the most to gain from artificial intelligence are the least protected by it—and that Africa bears a disproportionate share of the risk.
Somewhere in Lagos, a job applicant is being assessed by a hiring algorithm that has never been trained on a face that looks like hers. Somewhere in Nairobi, a healthcare AI system is making resource allocation recommendations based on data collected overwhelmingly in American hospitals. Somewhere in Abidjan, a farmer is trying to access a digital financial service that has no interface in his language and no regulatory framework in his country to protect him if it goes wrong. These are not hypothetical scenarios. They are the documented, measurable consequences of a global AI ecosystem that was built, largely, for someone else – and which the Global Center on AI Governance’s landmark research series has now mapped, country by country, with precision.
The Human Rights and AI:
A Global Index on Responsible AI Analytical Report Series, authored by Selamawit Abdella and published in February 2025, builds upon the first edition of the Global Index on Responsible AI – the GIRAI – which assessed 138 countries and jurisdictions across a comprehensive set of indicators relating to responsible AI governance. The human rights dimension that this brief series examines is one of three core components of the Index, and it asks a question that is simultaneously simple and structurally enormous: are countries actually protecting people from the harms that AI can cause, and providing them with a fair chance to benefit from what it can offer? The answer, for a significant proportion of the world’s population, is no.
“The contexts in which AI systems are being trained determine which lived experiences are being considered, and consequently, affect for whom the benefits of these technologies are accessible.” – Selamawit Abdella, GIRAI Report Series
The Geography of the AI Divide
The report’s most immediate finding is also its most politically uncomfortable: countries in North America and Europe consistently score highest in the human rights dimension of the GIRAI, while Africa and the Caribbean have the lowest average and median scores across the entire Index. This is not a coincidence or an anomaly. It is, the report argues, the contemporary expression of colonial histories — the continuation, in the infrastructure of artificial intelligence, of the same hierarchies of knowledge, resource and institutional power that structured the global order long before a single line of neural network code was written.
Africa is particularly exposed. Of the 41 African countries represented in the GIRAI — itself a notable achievement of the Index’s commitment to Global South research representation — Nigeria’s score in the human rights dimension stands at approximately 8.59 out of a possible 100. The country scores particularly low on gender equality, children’s rights and cultural and linguistic diversity within the AI context. The pattern holds across the continent. Countries including Chad, Eritrea, Liberia, Sierra Leone and the Central African Republic register near-zero scores across most thematic areas. In the language of the Index, these are ‘Very Low’ performers — not because their populations are less deserving of protection, but because the institutional infrastructure, legislative frameworks and non-state actor engagement needed to protect them from AI-related harms simply does not yet exist at adequate scale.
The implications extend well beyond the abstract. When a country has no data protection framework governing how its citizens’ information is collected and used by AI systems, those citizens have no legal recourse when that data is misused. When a country has no regulatory mechanism covering AI-driven hiring or credit-scoring systems, the structural biases encoded in those systems are applied without constraint. And when the AI systems themselves have been trained predominantly on data from the Global North, the experiences, languages, faces and needs of African populations are, at best, underrepresented and, at worst, actively distorted.
The Thematic Picture: Where the Risks Are Highest
The report examines seven thematic areas within the human rights dimension of the GIRAI: gender equality, data protection and privacy, public participation and awareness, bias and unfair discrimination, children’s rights, labour protection and the right to work, and cultural and linguistic diversity. The findings across these areas paint a picture of structural vulnerability that is, in some respects, more alarming than the overall regional scores suggest.
Gender equality and children’s rights are the two thematic areas with the lowest global performance – and in the Global South, the scores in these categories are, as the report states, ‘alarmingly low.’ This matters because the populations most likely to be harmed by biased AI systems – women in workplace algorithms, girls in educational platforms, children in content recommendation systems – are precisely the populations whose protection frameworks are most absent. Research cited in the report demonstrates that AI text-to-image generators systematically render women invisible in professional contexts, while facial recognition systems have well-documented failure rates for Black faces. These are not fringe failure modes. They are systemic outputs of systems trained on non-representative data, deployed in markets without the regulatory capacity to challenge them.
Cultural and linguistic diversity receives similarly scant attention, particularly in the Global South – a finding that carries significant commercial and social weight in a continent as linguistically rich as Africa. The rapid proliferation of Large Language Models built predominantly on English-language data creates AI systems that are, at a fundamental level, less accurate, less contextually appropriate and less useful when deployed in Hausa, Yoruba, Swahili, Amharic or any of the hundreds of other languages through which the majority of Africans navigate daily life. The risk is not only one of cultural exclusion. It is one of differential harm: people who engage with AI systems in environments the systems were not designed for receive worse outcomes, and have fewer mechanisms to challenge those outcomes.
Privacy and data protection is the one area of relative strength across all regions – suggesting that concerns about personal security and surveillance have penetrated policymaking frameworks in a way that more diffuse human rights considerations have not. But even this finding contains a caveat: having a data protection law on paper is not the same as having the institutional capacity to enforce it. The report’s correlation between rule of law scores and GIRAI human rights dimension performance makes this point with uncomfortable clarity.
Countries with strong adherence to the rule of law – specifically regulatory enforcement and civil justice – consistently outperform others in AI human rights protection. The law matters less than whether anyone enforces it.
Income, Development and the Compounding of Disadvantage
The report’s socioeconomic analysis is, in some ways, the most structurally significant finding in the entire series. Countries with higher incomes have substantially higher human rights dimension scores in the GIRAI, and this performance declines steadily as income decreases. The correlation is not perfect – there is wide disparity among high-income countries, indicating that money alone is not sufficient for rights-protective AI governance. Expertise, political commitment and institutional capacity must accompany financial resources. But the direction of the finding is unambiguous: people living in lower-income countries are less protected from AI-related harms, and this disadvantage compounds the economic inequality they already face.
The human development analysis – which uses the UNDP Human Development Index to map the same territory – reflects an identical pattern. Countries with lower human development scores have lower AI human rights governance performance. This creates what the report describes as a potential ‘double burden’ for vulnerable populations in already marginalised regions: they face both the regional disparity in AI governance and the specific vulnerabilities that lower development levels impose. In practical terms, it means that a woman living in a low-income country with low institutional capacity has neither the system-level protection nor the individual resource to challenge an AI system that treats her unfairly.
What Is Being Done – and What Needs to Happen
The report is not without grounds for cautious optimism. It highlights the African Continental Artificial Intelligence Strategy as evidence of growing regional consciousness around responsible AI governance. It identifies capacity-building initiatives across Africa, Latin America and the Caribbean that are developing local expertise in responsible AI design and ethical governance frameworks. Academic institutions and civil society organisations appear throughout the GIRAI data as meaningful actors in filling the gaps that government frameworks have not yet reached – a finding that carries particular relevance for Nigeria, where the density and energy of the tech and civil society ecosystem is among the continent’s richest.
The report makes a clear series of demands on policymakers, AI developers and advocacy organisations. Governments in lower-income countries need support – technical, financial and institutional – to build the regulatory capacity that responsible AI governance requires. AI developers building systems that will be deployed in African or other Global South markets have an obligation to ensure those systems are trained on representative data, tested for bias in the relevant context, and built with the linguistic and cultural specificity that those markets require. And advocacy organisations – including the press – have a responsibility to make the invisible consequences of AI governance gaps visible to the publics who bear them.
For Nigeria and for Africa more broadly, the GIRAI report arrives at a moment of particular strategic importance. The continent is in the early stages of building the digital infrastructure – the data centres, the connectivity, the fintech rails, the regulatory frameworks – that will determine how AI is developed and deployed here for decades to come. The decisions made in the next five years about data sovereignty, AI ethics frameworks, language inclusion and regulatory enforcement will either replicate the Global North’s governance advantages or close the gap. The GIRAI has measured where the gap currently stands. What happens next is a political and institutional choice – and it belongs, ultimately, to the governments, businesses and citizens of the countries whose scores the Index records.
