A new paper maps occupational exposure to frontier AI across 141 countries. Tajikistan starts near the bottom, with a score of 0.222 and a rank in the 21st percentile. Then the authors add the income its workers earn abroad.
Personal remittances received were equal to 47.9% of Tajikistan’s GDP in 2024. In the latest bilateral matrix, which dates from 2021, Russia accounted for 79% of the covered inflows. Weighting Tajikistan’s score by the exposure of the countries supplying those remittances raises it to 0.300, around the 75th percentile.
Most exposure estimates treat a national labour market as though it stopped at the border. But income earned in Russia pays for consumption in Tajikistan, so a change in the jobs migrants do can reach households whose own work has little direct exposure to current models.
The adjustment is still a hypothesis. The paper does not observe AI changing migrant employment, wages, remittances or household income, and it assigns the same occupational exposure to a job in every country. It identifies the channel and does not measure what moves through it.
For countries that depend heavily on remittances, the next labour-market study has to follow the workers abroad.
Assorted links
-
At a March workshop in Nairobi, government representatives and experts from twelve African countries repeatedly raised infrastructure, connectivity and access to compute as constraints on AI governance. The resulting OECD case study also argues that existing data-protection, consumer-protection and sectoral rules may be a more practical starting point than a new comprehensive AI law.
-
A review of 91 highly cited public-administration papers about AI found that 55% left the system they studied underspecified, 31% invoked a different type of system from the one they examined, and 41% drew conclusions broader than their evidence supported. The paper is under review. Its proposed remedy starts with naming what kind of system the research is actually about.