Generative AI is arriving in Korean offices faster than the rules that govern it
A Bank of Korea analysis estimated that around 3.4 million Korean jobs sit in the high-exposure zone for AI substitution, and the country's first comprehensive AI law took effect in January 2026 — but the workplace transition is running ahead of both statistics and statute.
Automation anxiety in Korea used to belong to the factory floor. The country operates one of the world’s densest industrial-robot fleets, and for decades the workers displaced by machines were welders and assemblers. Generative AI reverses the direction of exposure: the occupations most affected are the desk jobs — drafting, translation, coding, analysis, customer correspondence — that Korea’s education system spent a generation preparing its graduates for.
The scale is not trivial. In a 2023 analysis, Bank of Korea researchers estimated that roughly 3.4 million jobs, around 12 percent of employment, sat in the highest-exposure band for AI substitution, with white-collar and high-wage occupations disproportionately represented. The study’s authors were careful about the word “substitution” — exposure can mean augmentation as easily as replacement — but the inversion of the old pattern was the point: this time, the technology climbs the credential ladder rather than undercutting it.
Korean firms have moved quickly, in their fashion. The large conglomerates built internal AI platforms and restricted employee use of external chatbots over data-leakage concerns, then reintroduced the same capabilities inside their firewalls. Startups and marketing, gaming and software firms adopted the tools with fewer reservations. In between sits the long tail of small enterprises that employ most Koreans, where adoption is uneven and often informal — an employee quietly using a chatbot for correspondence the employer never sees.
The legal framework has been catching up. The AI Framework Act, passed by the National Assembly in December 2024 and effective from January 2026, made Korea one of the first jurisdictions with a comprehensive AI statute, taking a promotion-first approach with transparency obligations concentrated on high-impact and generative systems. What the law deliberately does not do is regulate AI’s labour-market consequences: questions of displacement, retraining and the terms on which AI enters the workplace are left to employment policy and, in practice, to firms.
That leaves the workplace transition governed by older instruments — an employment-insurance system built for factory closures, retraining programmes whose curricula age quickly, and a labour law centred on dismissal protection rather than task redesign. Labour researchers in Korea have pointed out the mismatch: the likeliest near-term outcome is not mass layoffs but a quiet thinning of entry-level white-collar work, the rungs by which juniors once learned the job. An economy already worried about how its young people start careers has reason to watch that rung disappear.
Korea enters this transition with real advantages — digital infrastructure, an adaptable workforce, and now a legal scaffold. Whether those translate into augmented workers rather than absent ones will be decided less by the statute than by thousands of unrecorded decisions in offices over the next few years.