AI Risk and Poverty: The Countries That Face Both - With No Safety Net

The Standard Assumption - and Why It Partially Fails

The standard assumption in AI disruption discourse is that wealthy countries absorb the impact first. The reasoning is sound as far as it goes: AI deployment requires digital infrastructure, enterprise software budgets, and a workforce already operating through digital systems. All of these correlate with income. Therefore, wealthy countries face disruption first.

This is largely true. Europe's 4.99/10 regional average is the highest in the world, and Africa's 3.37/10 is the lowest. Most of the countries with poverty rates above 50% score below 3.5/10 on AI exposure. The pattern holds.

But the pattern has exceptions that matter - not because they are common, but because they describe the most dangerous situation: significant AI exposure combined with extreme poverty and zero capacity to respond.

The High-Poverty, Non-Trivial AI Exposure Countries

Source: WorldJobsData AI exposure from ILO ILOSTAT (CC BY 4.0). Poverty rate: World Bank poverty headcount ratio at $3.65/day (2017 PPP), most recent available year per country from World Bank Open Data (CC BY 4.0). Risk velocity from WorldJobsData.

South Sudan: The Outlier That Explains the Pattern

South Sudan stands out. With a poverty rate of 76.5% and an AI exposure score of 4.13/10, it breaks the standard poverty-low-AI-exposure correlation. The reason is South Sudan's unusual economic structure.

South Sudan's formal economy is almost entirely built around oil extraction and government administration. The oil sector generates enormous revenues that flow through government ministries, international contractors, and NGO operations - all of which create clerical, administrative and coordination roles that score high on AI exposure. South Sudan has a larger formal administrative sector relative to its workforce than most of its neighbours, precisely because international aid operations and oil company contracts require documentation-intensive management.

The 4.13/10 score reflects that administrative class. The 76.5% poverty rate reflects the rest of the population - subsistence farmers and herders in the countryside who neither benefit from the formal economy nor show up in its AI exposure calculation. South Sudan does not have a workforce that is uniformly exposed to AI at 4.13/10. It has two separate workforces: a small formal sector with high AI exposure, and a large subsistence sector with almost none. The national average obscures this.

Turkmenistan: The Highest-Poverty High-AI Country

Turkmenistan is the clearest case of the double burden. At 4.74/10 AI exposure - in the same range as Saudi Arabia, Italy and Japan - and a 35.1% poverty rate, it is the highest-AI-exposure country with a significant poverty problem. Turkmenistan's 1.4 million formal workers are concentrated in government administration and the state energy sector, both of which create large clerical and managerial workforces. The country has a heavily centralized, bureaucratic economy - legacy of Soviet-era industrial planning - that generates administrative employment at a scale disproportionate to its GDP.

Turkmenistan's workers face a specific combination: exposure to AI disruption in their formal roles, with no labour market flexibility (the economy is state-controlled), no private-sector reskilling pathway, and a political system that does not allow independent worker organizing or advocacy.

DR Congo: Scale and Poverty Together

DR Congo is the case where the numbers are largest. With 30.6 million workers, an 85.3% poverty rate, and an AI exposure score of 3.35/10, it sits in a different quadrant than Turkmenistan: lower AI exposure, extreme poverty, enormous workforce. The 3.35/10 score reflects a workforce dominated by subsistence agriculture and artisanal mining - work that AI cannot currently automate economically.

But DR Congo is also growing, urbanizing, and developing a mobile money ecosystem that is formalizing economic activity at speed. Kinshasa - with a population of approximately 17 million - has a growing administrative and service class. The city's AI exposure is materially higher than the national average. As Congo urbanizes, the national score will rise. The question is whether governance, social investment, and economic opportunity keep pace with the formalization - and with the AI disruption that formalization brings.

What Separates Survivable Disruption from Catastrophic Disruption

AI disruption in a wealthy country - the Netherlands at 5.44/10, Denmark at 5.13/10 - arrives in an economy with unemployment insurance, reskilling programs funded by payroll taxes, a private education system that responds to labour demand, and political institutions capable of debating and implementing policy responses. The disruption is real and painful. It is also manageable.

AI disruption in a country with 76% poverty - even if the AI exposure score is lower - arrives in an economy where:

• There is no unemployment insurance for formal workers who lose their jobs
• There is no reskilling infrastructure - vocational training systems are absent or dysfunctional
• There is no economic buffer - workers who lose formal employment have no savings, no credit, and no safety net beyond family networks
• The political system may have no capacity or incentive to respond to worker displacement at scale
• There are no alternative formal sectors to absorb displaced workers

The absolute AI exposure score is not the only measure of risk. The capacity to absorb disruption is equally important. A country with a 3.5/10 AI exposure score and 80% poverty faces a version of AI disruption that could be more socially destabilizing than a country with a 5.0/10 score and a functioning welfare state.

The Most Important Question No One Is Asking

The current debate about AI and work is concentrated almost entirely on wealthy countries. Policy responses - AI governance frameworks, reskilling programs, minimum income proposals - are being developed in the US, EU, UK and other high-income economies. The implicit assumption is that developing countries have time to figure this out later.

The data suggests this assumption underestimates the speed and globalness of the transition. The AI tools being deployed in London and Amsterdam are the same ones available in Lagos and Nairobi. The clerical workers in Kinshasa face the same technological pressure as clerical workers in Warsaw. The difference is not the technology. It is the absence of institutions that can cushion the impact.

The countries facing the double burden - non-trivial AI exposure combined with extreme poverty and no safety net - are the ones where AI disruption could have the most severe human consequences. They are not yet in the center of the policy conversation. They should be.