## 1. AI-Related Policies and Regulation Initiatives in Korea

## Source details

**Canonical URL:** [1. AI-Related Policies and Regulation Initiatives in Korea](https://www.imf.org/-/media/files/publications/selected-issues-papers/2025/english/sipea2025013.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/selected-issues-papers/2025/english/sipea2025013.pdf.md)
- [Structured JSON version](/-/media/files/publications/selected-issues-papers/2025/english/sipea2025013.pdf.json)

---

### A. Scope, context, and objectives
- AI defined as technologies enabling machines to mimic human cognitive abilities (Cazzaniga et al., 2024); recent advances include generative AI (GenAI) such as large language models and generative pretrained transformers.
- Motivation for study:
  - Korea is a rapidly aging economy; AI can help address labor shortages, productivity slowdowns, and pressures on healthcare and pension systems, but poses risks of job displacement, reduced worker incomes, and increased inequality—especially for older workers.
  - Korea is a leading innovator and semiconductor producer: Korean semiconductor exports in 2024H1 accounted for "about 23 percent of global chips exports."
- Paper objectives and structure:
  - Section 2: stylized facts on AI adoption in Korean firms and workers.
  - Section 3: labor market impacts (complementarity and displacement).
  - Section 4: productivity and output impacts (model-based estimates and firm-level empirical analysis).
  - Section 5: Korea’s readiness for AI transition.
  - Section 6: policy recommendations.

### B. AI adoption — incidence, drivers, and firm patterns
- Key adoption statistics:
  - IBM 2023 (large enterprises): "42 percent" deploying AI; "40 percent" exploring AI; Korea-specific: "40 percent" of IT Professionals at large Korean enterprises reported using AI and "48 percent" reported active exploration in November 2023; only "6 percent" reported not using or exploring AI (sample average "15 percent").
  - Microsoft 2024 Work Trend Index: "73 percent" of Korean knowledge workers are utilizing AI at work; "About 80 percent" of Korean workers are bringing their own AI tools to work; "80 percent" of Korean business leaders believe companies need AI to stay competitive; "70 percent" indicate they would not hire candidates without AI skills.
  - Korea firm-level AI adoption (Survey of Business Activities): "1.4 percent in 2017" → "4.3 percent in 2022."
- Adoption patterns:
  - Higher AI adoption among larger firms (assets above the "75th percentile"), younger firms (age below "5 years"), and tech-intensive/innovative firms (patents, R&D expenditure per employee, intangible asset per employee).
  - Primary uses: product development; manufacturing; sales and marketing strategy; organizational management.
- Empirical adoption regression (linear probability model):
  - Dependent variable: AI usage (0-1).
  - Explanatory variables: firm size, age, digital capabilities (number of digital technologies excluding AI), innovativeness indicators, year and 2-digit industry fixed effects.
  - Result summary: Larger, younger, tech-related, and innovative firms are more likely to adopt AI; results robust (significance markers: *** p<0.01, ** p<0.05, * p<0.1).

### C. Labor market exposure, complementarity, and heterogeneity
- Occupational exposure framework based on Felten et al. (2021, 2023) and Pizzinelli et al. (2023): categories are "high exposure, high complementarity" (HEHC); "high exposure, low complementarity" (HELC); and "low exposure".
- Aggregate exposure statistics:
  - "Approximately 50 percent of employment falls within high-exposure occupations," split into "24 percent in high-complementarity and 27 percent in low-complementarity."
  - Korea’s proportion of high-exposure occupations is "slightly lower compared to some other advanced economies (AEs)."
- Distributional findings by occupation and demographics:
  - Professional occupations: large share in HEHC — likely to benefit from AI.
  - Clerical occupations: concentrated in HELC — greater displacement risk.
  - Gender: Women are more likely to work in high-exposure occupations than men; exposure is "roughly evenly split between low and high-complementarity."
  - Education: Higher education correlates with larger shares in high-exposure, especially high-complementarity roles.
  - Age: Younger workers more likely in high-exposure occupations than older workers.
  - Income deciles: Share in high-exposure occupations rises gradually with earnings deciles; high-complementarity jobs concentrated among upper-income groups.
- Worker reallocation dynamics:
  - Historical transitions suggest some mobility into higher-complementarity roles, but rigidities exist; microdata from the Korean Labor & Income Panel Study used to analyze transitions (2009–2022).

### D. Occupational transitions, life-cycle mobility, and reemployment
- Transition and mobility patterns:
  - Transition probabilities stable over 2009~2022, indicating a rigid labor market.
  - "31 percent" of those leaving low-complementarity jobs transition to roles with higher AI complementarity.
  - Female workers more likely than male workers to transition into HEHC roles.
  - Non-college-educated workers predominately in low-AI-exposure roles and less likely to move into high-complementarity positions.
- Life-cycle and age patterns:
  - Limited mobility between low- and high-complementarity jobs across the life cycle.
  - No-college group: share in low-exposure jobs rises with age; share in high-exposure declines.
  - College-educated group: HELC share decreases after age 50 with an increase in low-exposure roles.
  - Interpretation: older workers tend to move into simpler, repetitive jobs prior to or after retirement.
- Reemployment probabilities:
  - Older workers have lower likelihood of reemployment after unemployment.
  - Examples:
    - "35 percent" of older workers previously in HEHC roles found jobs in the same category within one year, versus "43 percent" for prime-age former HEHC workers.
    - Almost all older former HELC workers struggled to find high-exposure roles; only "25 percent" found jobs in low-exposure categories.
    - "31 percent" of prime-age former HELC workers found reemployment in the same category; "19 percent" transitioned into HEHC roles.
  - Gender: men have higher overall reemployment rates; women relatively more likely to move into HEHC roles.

### E. Model scenarios and macroeconomic impacts
- Task-based model (Cazzaniga et al., 2024) channels: (1) labor displacement; (2) labor complementarity; (3) overall productivity increase.
- Model calibration: uses estimated AI exposure and complementarity of Korean jobs.
- Three scenarios:
  - Scenario 1: labor displacement + labor complementarity.
  - Scenario 2: labor displacement + overall productivity increase.
  - Scenario 3: labor displacement + labor complementarity + overall productivity increase.
- Impact on TFP and output by scenario (left chart of Figure 13):
  - Scenario 1: "1.1 percent" increase in total factor productivity (TFP) and "8.4 percent" increase in output.
  - Scenario 2: "2.1 percent" increase in TFP and "4.2 percent" increase in output.
  - Scenario 3: "3.2 percent" increase in TFP and "12.6 percent" increase in output.
- Comparative literature examples (United States over the next decade):
  - Briggs and Kodnani (2023) projects "9.2 percent" increase in TFP.
  - Cazzaniga, et al. (2024) expects TFP impact within the range of "1.3 to 3.9 percent."
  - Acemoglu (2024) predicts TFP effects in the range of "0.53 to 0.66 percent."
- Caveat: timing and magnitude of AI impacts remain uncertain.

### F. Demographics, aging, and AI as an offset
- Demographic projection:
  - Using UN population projections and constant labor shares/participation rates, demographic aging would translate into a "16.5 percent" decline in output from 2023 to 2050.
- AI offset potential:
  - AI adoption could largely offset aging-induced output decline, particularly under Scenario 3 where AI operates through all three channels.

### G. Firm-level empirical evidence: productivity and profitability
- Patterns by firm percentile within industries (SNA A38):
  - AI usage concentrated among firms with higher productivity and profitability when classified into five percentile groups.
  - In 2022, AI usage among firms in the bottom "10 percent" of productivity and profitability distributions is higher than in the "10-40" and "40-60" percentiles, possibly reflecting AI start-ups among new entrants.
- Baseline regression (Survey of Business Activities, 2017–2022):
  - Dependent variables: firm productivity (revenue per employee) or firm profitability (net income per employee).
  - Explanatory variables: AI use (dummy), asset size, firm age; controls: 2-digit industry and year fixed effects; includes lagged (t-1) dependent variable.
  - Results: AI adoption has no significant impact on firm productivity but has a positive effect on firm profitability (Figure 15).
  - Larger firms (above the "75th percentile" by asset size) tend to have significantly higher productivity and profitability.
- Heterogeneity by firm characteristics and industry:
  - Including R&D expense and intangible assets and interactions with firm size/age:
    - Productivity benefits of AI evident only in larger, more mature firms.
    - Profitability gains especially clear for mature firms.
  - Industry-level:
    - Productivity improvements in larger firms within Professional, Scientific, and Technical Services, and Construction.
    - Mature firms in Information and Communication show substantial productivity gains.
  - Implication: early-stage AI adoption is not yet broad-based and could widen productivity and income gaps across firms and industries.

### H. Stylized global demand for semiconductors and Korea implications
- Industrial consensus forecasts expect AI-related semiconductor demand to double sales of chips to "1 trillion" by 2030 (Figure 18).
- If Korea maintains current market share (noted as "about 20 percent" of global semiconductor sales in this stylized analysis), this implies a significant boost in Korean semiconductor exports over the medium term.
- Note: earlier empirical fact—Korean semiconductor exports in 2024H1 accounted for "about 23 percent of global chips exports."

### I. AI preparedness: strengths and areas for improvement
- AI Preparedness Index (AIPI) dimensions: (1) digital infrastructure; (2) innovation and economic integration; (3) human capital and labor market policies; (4) regulation and ethics.
- Aggregate and dimensional findings:
  - Korea outperforms the median of advanced economies (AEs) in the aggregate AIPI and in three of the four dimensions (Figure 19).
  - Strengths:
    - Ranks third globally in “Innovation and Integration.”
    - High R&D investment, leadership in AI-related scientific research, strong global integration.
    - “Regulation and Ethics” preparedness exceeds the AE median (reflecting higher government effectiveness).
    - “Digital infrastructure” preparedness comparable to AE median; outstanding telecom infrastructure, mature e-commerce, robust public online services.
  - Areas for improvement:
    - Affordability and security of internet access (affordability measured by internet cost as percent of Gross National Income per capita, where Korea is higher than the AE median).
    - Overall “Human Capital and Labor Market Policies” slightly below the AE median:
      - Scope to increase public education spending and strengthen digital skills.
      - Need to enhance social protection and increase labor market flexibility to address AI displacement and adaptation needs.
  - Caveat: affordability metric has measurement limitations (Figure 20).

### J. Policy initiatives, legislation, and human capital actions
- National and legislative actions:
  - National AI committee and strategic directions include expanding AI infrastructure (including establishing a National AI Computing Center), incentivizing private investment, promoting broader AI adoption across sectors, risk management measures, and enacting the AI Basic Law.
  - Parliamentary proposals for AI Basic Law include establishment of an AI Committee, promotion of AI technology and data utilization, and formulation of AI ethical principles for reliability and safety.
- Human capital and labor market measures:
  - K-Digital Training (KDT) program: vocational, project-based training linking educational institutions, companies, and universities to provide real-world AI training.
  - Dynamic Economy Roadmap (July 2024) plans to meet AI professional demand by fostering domestic talent, attracting foreign experts, expanding specialized universities, accrediting in-house graduate programs, and exploring special visa/permanent residency measures.
- Social protection and labor market policy recommendations:
  - Enhance labor market flexibility.
  - Fund targeted training and reskilling programs (sector-based training, apprenticeships).
  - Strengthen social safety net via fiscal measures:
    - Expand unemployment insurance to more non-regular workers and the self-employed.
    - Enhance access and generosity of social assistance programs.

### K. Distributional challenges, sectoral opportunities, and priorities
- Distributional exposure:
  - Staff estimates that about "half of jobs are highly exposed to AI."
  - Groups simultaneously more exposed and more likely to benefit include women, young, high-skill and high-income workers; however, they also face greater potential adverse effects.
  - High labor market duality poses difficulties for worker mobility, especially for elderly groups.
- Sectoral and firm-level opportunity:
  - Korea’s leadership in semiconductors and increasing AI adoption position it to benefit from global AI demand.
  - Firm-level heterogeneity suggests large, mature firms gain most from early AI adoption; SMEs may lag, potentially widening productivity gaps.
- Priorities to harness AI while sharing benefits:
  - Increase investment in AI innovation and integration.
  - Advance regulatory frameworks (including AI Basic Law and ethical principles).
  - Enhance labor market flexibility.
  - Strengthen social safety net (expand unemployment insurance coverage; enhance social assistance).
  - Implement targeted training and reskilling programs (sector-based training, apprenticeships, KDT).
  - Attract foreign AI professionals through visa and residency measures.

*Source: sipea2025013 - 1. AI-Related Policies and Regulation Initiatives in Korea (International Monetary Fund, January 21, 2025).*

### 1. AI-Related Policies and Regulation Initiatives in Korea _____________________________ 21

### 1. AI-Related Policies and Regulation Initiatives in Korea _____________________________ 21

### A. Introduction — scope and context
- AI defined as technologies enabling machines to mimic human cognitive abilities (Cazzaniga et al., 2024).
- Recent advances include generative AI (GenAI) such as large language models and generative pretrained transformers.
- IBM 2023 survey (large enterprises, over 1,000 employees): "42 percent of surveyed IT Professionals deploying AI and an additional 40 percent reporting active exploration in November 2023."
- IMF WEO, April2024 documents AI performance gains across cognitive domains and cites generative pretrained transformers as a milestone (GRE mathematics test benchmark context included).

Key contextual observations:
- Korea is a rapidly aging economy; AI can help address labor shortages, productivity slowdowns, and pressures on healthcare and pension systems, but poses risks of job displacement, reduced worker incomes, and increased inequality—especially for older workers.
- Korea is a leading innovator and semiconductor producer; Korean semiconductor exports in 2024H1 accounted for "about 23 percent of global chips exports."
- Literature coverage: growing global and Korea-specific work, but comprehensive country-level assessments of AI’s broad economic and policy implications for Korea are limited.

Paper objectives and structure:
- Present stylized facts on AI adoption in Korean firms and workers (Section 2).
- Examine labor market impacts including complementarity and displacement using Korean labor survey data (Section 3).
- Assess productivity and output impacts via model-based estimates and firm-level empirical analysis (Section 4).
- Analyze Korea’s readiness for AI transition and areas for improvement relative to global leaders (Section 5).
- Conclude with policy recommendations (Section 6).

### B. AI Adoption — incidence and drivers
Findings on adoption and use:
- IBM 2023: In Korea, "40 percent of IT Professionals at large Korean enterprises reported using AI" and "48 percent reported active exploration in November 2023"; only "6 percent" reported not using or exploring AI (sample average "15 percent").
- Microsoft 2024 Work Trend Index: "73 percent of Korean knowledge workers are utilizing AI at work" versus global average "75 percent"; "About 80 percent of Koreans workers are bringing their own AI tools to work" (global "78 percent"); "80 percent of Korean business leaders believe that their companies need to adopt AI to stay competitive" and "70 percent indicate they would not hire candidates without AI skills."
- Korea Statistics, Survey of Business Activities: AI user share among Korean firms rose from "1.4 percent in 2017 to 4.3 percent in 2022."
- Highest AI adoption by industry: information and communication industry (ICT) with "about 18 percent in 2022", followed by professional services.

Patterns of adoption:
- Higher AI adoption among:
  - Larger firms (assets above the "75th percentile"),
  - Younger firms (age below "5 years"),
  - Tech-intensive firms (patents, active exploration of technologies).
- Firms mainly use AI for product development, then manufacturing, sales, marketing strategy, and organizational management.

Empirical analysis: adoption regression
- Method: linear probability model with AI usage (0-1) as dependent variable; explanatory variables include firm size, age, digital capabilities (number of digital technologies excluding AI), innovativeness (patent ownership, R&D expenditure per employee, intangible asset per employee), year and 2-digit industry fixed effects.
- Result summary (Figure 5): Larger, younger, tech-related, and innovative firms are more likely to adopt AI; results robust across specifications.
- Reported statistical annotation: significance markers used (*** p<0.01, ** p<0.05, * p<0.1).

### C. AI and the labor market — exposure, complementarity, and heterogeneity
Occupation exposure and complementarity framework:
- Builds on Felten et al. (2021, 2023) for "exposure" and Pizzinelli et al. (2023) for "potential AI complementarity".
- Occupational classification: "high exposure, high complementarity"; "high exposure, low complementarity"; and "low exposure".

Aggregate exposure statistics:
- "Approximately 50 percent of employment falls within high-exposure occupations," split into "24 percent in high-complementarity and 27 percent in low-complementarity."
- Korea’s proportion of high-exposure occupations is "slightly lower compared to some other advanced economies (AEs)."

Occupational distributional findings:
- Professional occupations: large share in "high exposure, high complementarity" — likely to benefit from AI.
- Clerical occupations: concentrated in "high exposure, low complementarity" — at greater risk of displacement.
- Net implication: Korea may experience a polarized impact with both risks (displacement, negative income effects for low-complementarity roles) and opportunities (growth in high-complementarity roles).

Demographic and income heterogeneity:
- Gender:
  - Women are more likely to work in high-exposure occupations than men.
  - That exposure is "roughly evenly split between low and high-complementarity", implying both greater risks and greater opportunities for women.
- Education:
  - Higher education levels correspond to larger shares in high-exposure occupations, especially those with high complementarity.
  - Suggests AI may impact highly skilled workers more than prior technologies; higher exposure is offset by greater complementarity potential.
- Age:
  - Younger workers are more likely to be in high-exposure occupations than older workers, largely due to higher education levels.
- Income deciles:
  - Share of employment in high-exposure occupations rises gradually with earnings deciles.
  - Jobs with high potential for AI complementarity are more concentrated among upper-income groups.

Worker reallocation dynamics:
- Historical patterns of job transitions provide insights on adaptation: over time, workers may shift into roles with higher AI complementarity, but some may struggle to adapt.
- The analysis uses microdata from the Korean Labor & Income Panel Study to study transitions among occupations with varying exposure and complementarity (analysis introduced; detailed transition results follow in subsequent sections).

### Key figures and statistics (selected, exact)
- IBM 2023: "42 percent" deploying AI; "40 percent" exploring AI (IT Professionals, large enterprises).
- IBM 2023 enterprise use series shown historically: Oct-19, Apr-21, Apr-22, Apr-23, Nov-23 with category levels illustrated (percent shares reported in figures).
- Korea firm-level AI adoption: "1.4 percent in 2017" → "4.3 percent in 2022."
- Korean semiconductor exports in 2024H1: "about 23 percent of global chips exports."
- Microsoft 2024 Work Trend Index: "73 percent" of Korean knowledge workers use AI at work; "80 percent" bring own AI tools; "80 percent" of leaders say companies need AI; "70 percent" would not hire without AI skills.
- Employment exposure split: "50 percent" high-exposure overall; "24 percent" high-complementarity; "27 percent" low-complementarity.

*Source: sipea2025013 - 1. AI-Related Policies and Regulation Initiatives in Korea (International Monetary Fund, January 21, 2025).*

### 14.      College-educated individuals in AI-intensive jobs tend to remain in similar occupation

### 14.      College-educated individuals in AI-intensive jobs tend to remain in similar occupation

### Occupational transitions and mobility
- The transition probability between occupations shows a stable pattern without significant changes over time (2009~2022), suggesting a rigid labor market in Korea.
- 31 percent of those leaving low-complementarity jobs transition to roles with higher AI complementarity, indicating a potential job-ladder pathway.
- Female workers are more likely than male workers to transition into high-exposure, high-complementarity (HEHC) roles, regardless of previous job exposure level.
- Non-college-educated workers are predominantly in low-AI-exposure roles and are less likely to move into high-complementarity positions.
- Note: The bars in Figure 10 represent the average values from 2009 to 2022.

### Life-cycle profiles and age patterns
- Limited mobility between low- and high-complementarity jobs throughout the life cycle.
- No-college group: the proportion of workers in low-exposure jobs rises significantly with age, while the share in high-exposure jobs declines steadily.
- College-educated group: the proportion of HELC (high-exposure, low-complementarity) jobs decreases after age 50, with a corresponding increase in low-exposure roles.
- Interpretation: as workers age, they are more likely to transition into simpler, repetitive jobs, potentially due to a shift toward manual labor just before or after retirement.
- Note: Figure 11 plots estimated share of employment by age for each exposure category for college- and non-college-educated workers, according to calculations described in Cazzaniga, and others (2024).

### Reemployment probabilities and age/gender differences
- Older workers have lower likelihood of finding reemployment after being unemployed.
- Among those unemployed last year, prime-age workers generally secure new jobs within one year more easily.
  - Example: 35 percent of older workers who were previously in HEHC roles managed to find jobs in the same category, compared to 43 percent of prime-age former HEHC workers.
  - Almost all older former HELC workers struggled to find positions in high-exposure roles, with only 25 percent finding jobs in low-exposure categories.
  - In contrast, 31 percent of prime-age former HELC workers found reemployment in the same category, and 19 percent managed to transition into HEHC roles.
- Gender: men have higher overall reemployment rates, but women are relatively more likely to move into HEHC roles.
- Note: The bars in Figure 12 represent the average values from 2010 to 2022.

### Model estimates: AI channels and scenarios
- The task-based model (Cazzaniga et al., 2024) quantifies AI impact through three channels:
  1. labor displacement;
  2. labor complementarity;
  3. overall productivity increase.
- Model calibration uses estimated AI exposure and complementarity of jobs in Korea.
- Three scenarios for AI impacts:
  - Scenario 1: labor displacement + labor complementarity.
  - Scenario 2: labor displacement + overall productivity increase.
  - Scenario 3: labor displacement + labor complementarity + overall productivity increase.
- Impact on TFP and output by scenario (left chart of Figure 13):
  - Scenario 1: 1.1 percent increase in total factor productivity (TFP) and 8.4 percent increase in output.
  - Scenario 2: 2.1 percent increase in TFP and 4.2 percent increase in output.
  - Scenario 3: 3.2 percent increase in TFP and 12.6 percent increase in output.
- Comparative literature examples for the United States over the next decade:
  - Briggs and Kodnani (2023) projects 9.2 percent increase in TFP.
  - Cazzaniga, et al. (2024) expects TFP impact within the range of 1.3 to 3.9 percent.
  - Acemoglu (2024) predicts TFP effects in the range of 0.53 to 0.66 percent.
- Caveat: the impact and timing of AI on productivity and economic outcomes remain uncertain.

### Aging, demographics, and offset by AI
- Using UN projections of population trends and assuming labor share of income and labor force participation rates across gender and age groups remain constant, demographic aging would translate into a 16.5 percent decline in output from 2023 to 2050.
- AI adoption could largely offset aging-induced output decline, with the largest offset occurring in Scenario 3 where AI affects all three channels.

### Empirical analysis: AI usage and firm outcomes
- AI usage is more prevalent among firms with higher productivity and profitability when firms are classified into five percentile groups within industries (SNA A38).
  - Firm productivity measured as revenue per employee; profitability measured as net income per employee.
- In 2022, AI usage among firms in the bottom 10 percent of productivity and profitability distributions is higher than in the 10-40 and 40-60 percentiles, possibly reflecting higher presence of AI start-ups among new entrants.
- Baseline regression (Survey of Business Activities, 2017–2022):
  - Dependent variables: firm productivity or firm profitability.
  - Explanatory variables: AI use (dummy), asset size, firm age; controls: 2-digit industry and year fixed effects; includes lagged (t-1) dependent variable.
  - Results: AI adoption has no significant impact on firm productivity but has a positive effect on firm profitability (Figure 15).
  - Larger firms (above the 75th percentile by asset size) tend to have significantly higher productivity and profitability.

### Heterogeneity by firm characteristics and industry
- Including R&D expense and intangible assets (dummy variables) and interaction terms between AI use and firm size/age:
  - Productivity benefits of AI are evident only in larger, more mature firms.
  - Profitability gains from AI adoption are particularly clear for mature firms.
  - Implication: large and mature companies in Korea experience significant improvements in both productivity and profitability from AI.
- Industry-level heterogeneity (Figure 17):
  - Productivity improvements observed in larger firms within Professional, Scientific, and Technical Services, and Construction sectors.
  - Mature firms in Information and Communication show substantial productivity gains.
  - Conclusion: broad-based productivity increase is not yet evident in early stages of AI adoption; AI adoption could widen productivity and income gaps across firms and industries.

### Stylized analysis: global AI demand and semiconductors
- Industrial consensus forecasts expect AI-related semiconductor demand to double sales of chips to 1 trillion by 2030 (Figure 18).
- If Korea maintains current market share (about 20 percent of global semiconductor sales), this implies a significant boost in Korean semiconductor exports over the medium term.
- Uncertainty remains high regarding AI-driven global semiconductor demand.

### AI preparedness: strengths and areas for improvement
- The AI Preparedness Index (AIPI) from Cazzaniga et al. (2024) comprises four dimensions: (1) digital infrastructure, (2) innovation and economic integration, (3) human capital and labor market policies, (4) regulation and ethics.
- Korea outperforms the median of advanced economies (AEs) in the aggregate AIPI and in three of the four dimensions (Figure 19).
- Strengths:
  - Ranks third globally in “Innovation and Integration”.
  - High R&D investment, leadership in AI-related scientific research, strong global integration.
  - “Regulation and Ethics” preparedness exceeds the AE median, reflecting higher government effectiveness.
  - “Digital infrastructure” preparedness comparable to AE median; outstanding telecom infrastructure, mature e-commerce, robust public online services.
- Areas for improvement:
  - Affordability and security of internet access (affordability measured by internet cost as percent of Gross National Income per capita, where Korea is higher than the AE median).
  - Overall “Human Capital and Labor Market Policies” is slightly below the AE median.
    - Human capital sub-indicators: scope to increase public education spending and strengthen digital skills.
    - Labor market policies: need to enhance social protection (to address AI displacement effects) and increase labor market flexibility (to allow firms to better adapt to AI-driven changes).
- Note: Figure 20 details sub-categories of the AIPI; affordability metric caveat noted regarding measurement limitations.

*Source: sipea2025013*

### 29.      Active and ongoing policy efforts are being made to promote AI adoption while

### 29.      Active and ongoing policy efforts are being made to promote AI adoption while

### Policy initiatives and strategic directions
- National initiatives and the establishment of a national AI committee outline key directions for the country’s AI strategy, including:
  - expanding AI infrastructure (including establishing a National AI Computing Center),
  - incentivizing private investment in AI development through policy-backed financial support,
  - promoting broader AI adoption across a wide range of sectors,
  - implementing risk management measures, including enacting the AI Basic Law.
- Other focus areas include nurturing AI startups and talent, advancing core technologies and innovation, establishing a foundation for sustainable AI development, and developing legal principles for AI responsibility and rights.
- The Dynamic Economy Roadmap (announced in July 2024) includes plans to meet growing demand for AI professionals by:
  - fostering domestic talents and attracting foreign experts,
  - expanding specialized universities in high-tech fields,
  - accrediting in-house graduate programs,
  - exploring special visa options and ways to expedite the permanent residency and naturalization processes.

### Human capital, training, and labor market actions
- The government has introduced initiatives to train and transition professionals into AI-related fields, including the K-Digital Training (KDT) program:
  - KDT is a vocational training initiative that brings together educational institutions, companies, and universities to offer project-based, real-world training.
- Policies to strengthen the labor market and social protection recommended to support broad-based AI adoption include:
  - enhancing labor market flexibility,
  - funding targeted training and reskilling programs (including sector-based training and apprenticeships),
  - strengthening the social safety net through fiscal policy measures, specifically:
    - expanding unemployment insurance to more non-regular workers and the self-employed,
    - enhancing access and generosity of social assistance programs.

### Legislative and regulatory developments
- Parliamentary legislative proposals are in progress to enact the AI Basic Law. Key elements of these proposals include:
  - establishment of an AI Committee,
  - promotion of AI technology and data utilization,
  - formulation of AI ethical principles to ensure reliability and safety in development and use.
- Regulatory efforts also target minimizing negative side effects by establishing ethical standards and ensuring the trustworthiness of AI technology.

### Economic impact, projections, and sectoral implications
- Potential macroeconomic gains:
  - Under a scenario where AI complements job functions and increases overall productivity, AI adoption could boost productivity and output by about 3 percent and 13 percent over the next decades, respectively.
  - These gains could largely offset the estimated negative impact of population aging on output.
- Korea’s position and sectoral opportunity:
  - Korea is at the global forefront in AI adoption, with increasingly more companies and workers actively exploring the use of AI.
  - As a top semiconductor producer, Korea is set to benefit from the global AI boom: it is projected that global AI demand will lead to the doubling of Korean chips exports by 2030.
- Firm-level variation:
  - Adoption rates are higher among larger and younger firms with stronger technological capacities.
  - Firm-level analysis suggests productivity and output enhancements are significant only in large and mature Korean firms, which could further exacerbate existing productivity gaps between large firms and SMEs.

### Distributional challenges and exposure
- Labor market exposure and distributional effects:
  - Staff estimates that about half of jobs are highly exposed to AI.
  - While some groups may benefit more from AI adoption, others may be more likely to be adversely affected:
    - Women, young, high-skill and high-income groups are more likely to be adversely affected by AI, while at the same time may benefit more from AI adoption.
  - High labor market duality poses significant challenges for workers to switch jobs, especially for elderly groups.

### Policy recommendations and priorities
- To fully harness AI’s potential while ensuring widespread sharing of benefits, targeted policies are needed, including:
  - increasing investment in AI innovation and integration,
  - advancing adequate regulatory frameworks (including the AI Basic Law and ethical principles),
  - enhancing labor market flexibility,
  - strengthening the social safety net (expanding unemployment insurance coverage and enhancing social assistance),
  - implementing targeted training and reskilling programs (sector-based training, apprenticeships, and programs like KDT),
  - attracting foreign AI professionals through visa and residency measures to meet talent demand.

*Source: IMF Selected Issues Paper content unit sipea2025013*

---


_Source: https://www.imf.org/-/media/files/publications/selected-issues-papers/2025/english/sipea2025013.pdf_