Abstract
The labor-market effects of frontier AI are crucial for workers, firms, and policymakers, yet current evidence is primarily derived from a handful of high-income economies. The capabilities of frontier AI are uneven across work tasks, and national economies diverge in their labor allocation.
Key Findings
We introduce a national AI exposure metric combining occupation-level exposure scores and international employment data for 141 countries. Our findings indicate that high-income countries are significantly more exposed than low-income countries, with Europe and Central Asia being 50% more exposed than Sub-Saharan Africa. Additionally, we observe a gender gap, where women are more exposed than men in 91% of countries, driven by their concentration in white-collar and sales occupations. Exceptions exist in countries where women are primarily employed in agriculture and household enterprises.
Validation and Mechanism
We validate our national AI exposure estimates by demonstrating their predictive power regarding national AI adoption statistics published by Anthropic, Microsoft, and OpenAI. Furthermore, we identify a new mechanism for indirect exposure due to cross-country income dependencies. For instance, Tajikistan heavily relies on foreign workers remitting money home. Although Tajikistan's direct exposure to frontier AI is below average, its remittance-accounted exposure is above average due to 37% of its GDP stemming from Russian remittances, a country with high exposure.
Conclusion
Our research illustrates that the variation in national exposure is significant enough that policy responses tailored to U.S. or European labor markets may not be applicable elsewhere.