As of May 2026, employment insurance subscribers under the age of 30 numbered about 2,233,000, down 65,000, or 2.8%, from a year earlier. In the same month, total subscribers rose by 268,000 (1.7%). That means only the bracket of people in their 20s moved the other way while the total grew. The age group with the largest increase was those 60 and older, up 207,000 (7.5%).
Employment insurance enrollment is the indicator that shows most quickly how many people are working for wages. If the whole had shrunk together because of a weak economy the explanation would be simple, but what is happening now is closer to the opposite. The growth rate of employed people aged 15-29 was negative for five months straight from January through May this year, and this is the only age bracket that fell for five consecutive months.
Broken down by occupation, the decline is concentrated in particular areas. Among subscribers in their 20s in May, writers and translators fell 20.6%, accounting and bookkeeping clerks 11.5%, software developers 10.1%, designers 7.6% and legal clerks 6.1%. Writing sentences, balancing books, coding and turning out drafts are the areas where generative AI reaches deepest at the first-draft stage. Looking only at those under 30, information and communications fell 9.3% and professional, scientific and technical services fell 4.1%.
An analysis released by the Bank of Korea last October traced this trend over a longer period. In the roughly three years since ChatGPT appeared in November 2022, 211,000 jobs held by people aged 15-29 disappeared, while jobs held by people in their 50s rose by 209,000. Of the 211,000 youth jobs lost, 208,000 came from industries in the top half for AI exposure. The totals roughly traded places, but the ages of those who left and those who filled in were different.
Opposite signals came out one after another in the same week. Sam Altman, CEO of OpenAI, said at a conference in May this year that he had mostly been right on technology forecasts but considerably wrong on social and economic effects. Dario Amodei, CEO of Anthropic, who warned last May that AI could wipe out half of entry-level jobs, also explained that the point of the warning was to buy time to respond, and said he had no intention of prophesying doom. The share of CEOs who said AI investment would lead to large-scale workforce cuts fell from 46% in January last year to 20% in May this year.
The two trends do not clash. Meta cut 8,000 jobs in May and Amazon cut 16,000 as it had signaled, but the change that registers most in the statistics is not layoffs. At a seminar held in Seoul on the 7th by the Federation of Korean Industries (한국경제인협회), participants said employment changes from AI adoption are beginning to show up more as reduced new hiring than as cuts to existing staff. One business owner in his 60s who does website building and systems integration decided not to hire developers or designers this year, judging that his current staff is enough with generative AI.
Lee Chul-hee, a professor at Seoul National University, pointed out at the seminar a split in which occupations with high pay and skill requirements such as finance and law have higher AI exposure, while fields short of workers such as care work, restaurants and transportation are hard to replace. As AI takes over entry-level tasks, the very channel through which young people build up skills could grow thin. This year the certified public accountant exam saw the number of applicants and the competition ratio drop to five-year lows, and as major accounting firms cut entry-level hiring, more successful candidates were left without a firm to join. Legal clerk subscribers numbered 7,175 in May, down for 13 straight months.
Macro indicators point the same way. The employment elasticity, which shows how much employment growth produces, is estimated at 0.24 this year, the lowest since 0.13 in 2018. That is less than half of last year's 0.64.
In its May outlook, the KDI projected 2.5% growth this year and an increase of 170,000 in the number of employed people. A state in which the economy grows while hiring less is hardening into the statistics.
When 637 university students nationwide were asked last September, 82.1% said their future careers could be shaken by AI. At the University of Arizona commencement in May, former Google CEO Eric Schmidt drew boos when he said AI would become part of work, and at Stanford University's commencement in June, some 200 graduates held up placards and walked out of the hall during Google CEO Sundar Pichai's turn.
The Ministry of Finance and Economy (재정경제부) is expanding practical AI training for young people and programs linked with companies, and a joint measure from the ministries concerned will come out in the second half. What policy has to aim at is the width of the door into a first job. Machines have taken the draft work once handed to new hires, so where will people pick up skills they can do by hand? Which way the curve for those in their 20s bends in next May's employment insurance figures will stand in for the answer.
