35 Countries Have the Same AI Job Risk as the United States - Most Aren't Western

AI Does Not Care About GDP

The assumption embedded in most AI and work discussions is that AI disruption is a problem for wealthy countries first. The reasoning seems logical: wealthy countries have more AI infrastructure, more knowledge workers, more office-based employment. The assumption is partially right - wealthy countries do tend to score higher. But it breaks down at the level of individual countries, and it breaks down badly.

The US scores 5.07/10 on AI workforce exposure. That is the employment-weighted average across all 143.1 million US workers, calculated from ILO ILOSTAT (CC BY 4.0) employment data using ISCO-08 occupation groups, with AI exposure scores assessed by Claude (Anthropic) for each group. A score of 5.07 means the average US worker is in a job where AI can handle a moderate-to-significant share of tasks - not imminent replacement, but meaningful disruption.

Now consider Poland. GDP per capita: $28,420 - less than a third of the US. AI exposure score: 5.04/10 - statistically identical to the US. Or South Africa. GDP per capita: $6,598 - roughly 7% of the US level. AI exposure score: 4.84/10. Or Tuvalu, a Pacific microstate with a GDP per capita of $6,041. AI exposure score: 4.77/10.

These are not rounding errors. They reflect a genuine structural reality: when a country's workforce shifts toward formal clerical and service employment, its AI exposure rises - regardless of what the country earns per capita.

The Full Comparison: Countries Within Range of the US

Source: WorldJobsData from ILO ILOSTAT (CC BY 4.0) employment data and World Bank Open Data (CC BY 4.0). AI exposure scores weighted by employment share across ISCO-08 major groups. Data as of 2026.

The Mechanism: Clerical Work Is the Equaliser

The reason income-diverse countries converge on similar AI exposure scores is that one occupation group dominates the calculation: clerical support workers (ISCO-08 Group 4), which scores 8.5/10 - the highest of any major group in the dataset. Data entry, scheduling, correspondence, form processing, records management - these are tasks where AI has already demonstrated near-human competence, and where deployment is actively underway.

Countries with large clerical sectors score high regardless of income. Japan has 14.5 million clerical workers - 20.5% of its workforce. South Korea has 3.6 million (12.5%). South Africa has 1.96 million (15.4% of formal workers). Poland's large administrative class drives its 5.04/10 score. Hungary's public sector clerks drive its 4.82/10.

The inverse holds too. Countries that appear low-risk are often low-risk simply because they have small formal sectors. A country with 80% of workers in subsistence agriculture has almost no clerical workers - so its AI exposure score is low. But that is a statement about its economy, not a statement about AI's incapacity to affect it.

The Three Surprising Countries

Poland: Eastern Europe at US-Level Risk

Poland's 16.3 million workers score 5.04/10 on AI exposure - essentially equal to the US. Poland has undergone rapid economic formalization over the past 30 years, building a large administrative class, a significant business process outsourcing sector, and a professional workforce that mirrors Western Europe's structure. Its clerks, administrators and professionals face the same AI disruption timeline as their counterparts in Frankfurt or Chicago - with fewer economic buffers and a smaller social safety net.

South Africa: A Formal Economy Inside a Developing Country

South Africa's 4.84/10 score comes from its formal economy, which is structured around services, finance, retail administration and government - roles that score high on AI exposure. South Africa's apartheid-era economic history created a two-tier labour market: a large formal sector with Western-style occupation structures, sitting alongside a large informal and subsistence sector. The formal sector drives the AI exposure score. That 15.4% clerical share of formal workers - the highest in sub-Saharan Africa - is what puts South Africa in the same risk band as Japan, Germany and Australia.

Ukraine: Disruption Without Economic Cushion

Ukraine scores 4.63/10 on AI exposure with a GDP per capita of $5,866. It has a large educated professional and clerical workforce - legacy of Soviet-era universalism in higher education and bureaucratic employment. Those workers face US-level AI exposure with a fraction of US-level economic resilience, social protection, and reskilling infrastructure. The combination is the most challenging in the dataset.

What This Means: The Global AI Disruption Is Already Borderless

The implication of this data is not reassuring. AI disruption is not a problem that wealthy countries will experience first and then export solutions to poorer ones. It is a problem that arrives simultaneously across income levels, wherever formal office-based employment exists.

Poland's clerical workers face the same automation pressure as US clerical workers. South Africa's administrators face the same AI tools as UK administrators. The AI models available in Johannesburg are the same ones available in New York. The difference is not the technology. It is the capacity to absorb the disruption - reskilling programs, unemployment insurance, education system flexibility, and the economic dynamism to create new roles for displaced workers.

On that dimension, the 35 countries with US-level AI exposure do not all have US-level capacity to respond. That is the gap that matters.