Korea lost 11 top AI researchers a month over 13 years, World Bank says

Korea lost an average of 11 of its top AI researchers and inventors to other countries every month between December 2011 and December 2024, the World Bank said in a report released Tuesday. That works out to roughly 1,716 people over 13 years.

Korea ranked fifth for net outflow among the 21 countries the bank compared, behind India, which came first with 200 a month, Brazil with 19, Italy with 17 and France with 12.

The findings appear in "World Development Report 2026: The Promise of Artificial Intelligence", the World Bank Group's annual flagship study, which this year takes up what AI means for developing economies.

The bank drew on the database of Zeki Data, a British firm that tracks AI talent, and analyzed research papers, patents, code repositories and records of new model development. It classified about 658,000 people in the top 1 percent of contributors to AI research and development worldwide as top talent, then followed their movements.

"Zeki is a leading tracker of the world's best AI talent, and it works from actual research output and technical contribution," said Kim Yong-hee, a business administration professor at Sunmoon University. "The scale of Korea's AI talent outflow has now been confirmed in objective numbers."

The United States absorbed the most by a wide margin, with a net inflow of 197 people a month. Britain followed at 20 and Germany at 17. The bank attributed the gap to the concentration of the world's leading AI companies and research capacity in the United States, which has made Silicon Valley the hub that pulls in global AI talent.

Earlier studies point the same way. Stanford University's AI Index measured Korea's net AI talent migration at minus 0.30 per 100,000 people in 2023, or 0.30 more departures than arrivals. The figure widened to minus 0.36 in 2024. The National Information Society Agency's AI Index 2026 found net outflows in each of the four years through 2025. It put the blame on low pay and thin research infrastructure.

China was left out of the World Bank analysis for lack of data. The exclusion matters because Chinese labs have been drawing researchers trained abroad. Yang Zhilin, the founder of Moonshot AI, which built the open-source Kimi K3 model, holds a doctorate from Carnegie Mellon University, and his case set off its own brain-drain debate in the United States. 

"Had China been included, Korea's net outflow of AI talent could well have been larger," an AI industry source said.

The report also exposed a contradiction in Korea's AI base. The bank measured code uploaded to GitHub, the largest open-source code-sharing platform, from 2011 to 2025 and found Korea among the world's leaders in development activity relative to per capita GDP. Working developers are deeply engaged, even as the country's best researchers leave.

"It means we have succeeded at producing AI talent and failed at holding on to it," said Wi Jong-hyun, dean of Chung-Ang University's College of Virtual Convergence. "Unless research conditions and pay at Korean conglomerates and universities improve dramatically, the outflow is likely to get worse."

Much of the report argues that AI is more opportunity than threat for developing economies. A smaller share of their jobs can be automated, and using cheap AI tools rather than the capital-hungry frontier models could raise productivity and create work. In high-income countries, Korea among them, about 14.2 percent of jobs are exposed to automation by AI, against 4.5 percent in low- and middle-income countries.

The bank warned those countries against putting off AI investment. Indermit Gill, its chief economist, said today's developing economies missed the first Industrial Revolution and spent the next two centuries paying for it.

"They cannot afford to miss this one," Gill said.

BY PARK TAE-IN, LEE SEUNG-HO [cho.yongjun1@joongang.co.kr]