## Executive Summary ## Source details **Canonical URL:** [Executive Summary](https://www.imf.org/-/media/files/publications/imf-notes/2026/english/insea2026002.pdf) ## Other formats - [Markdown version](/-/media/files/publications/imf-notes/2026/english/insea2026002.pdf.md) - [Structured JSON version](/-/media/files/publications/imf-notes/2026/english/insea2026002.pdf.json) --- ### Context and purpose - Synthesizes insights from a high-level workshop and a scenario-planning exercise on the global economic and financial implications of artificial intelligence (AI), hosted by the IMF in collaboration with the Economics of Transformative AI Initiative (EconTAI) at the University of Virginia (December 10–11, 2025). - Framing: treat AI as a macro-critical transition rather than a standard technology shock because of its potential to restructure the global economy through task automation and accelerated research and development. - Emphasis: macroeconomic path shaped not only by frontier capability but by speed and breadth of diffusion and institutional/infrastructure readiness. ### Key near-term constraints (workshop consensus) - Physical and organizational bottlenecks can limit economy-wide productivity gains even with rapid frontier improvement: - energy and grid capacity - data-center infrastructure - persistence of tasks requiring physical presence - Labor market outcomes depend on changes in task composition and expertise content of work, not only on headline job displacement; this implies potentially divergent wage and employment effects across occupations. - Market structure and rents matter: scale economies in compute and data can amplify “winner-take-most” dynamics, raising the risk of greater within- and cross-country inequality. - Governance and social cohesion are critical constraints: institutions’ ability to acquire and maintain public trust determines the stability of the transition. ### Scenario-based assessment (five-year horizon) - Two alternative diffusion trajectories were explored assuming rapid technological capacity progress: - Baseline scenario: - Adoption is uneven and constrained by regulatory, infrastructure, and organizational frictions. - Runaway diffusion scenario: - Adoption accelerates broadly, with rapid expansion of automation across services and industry, and more front‑loaded labor displacement. - Common projection themes across scenarios: - Productivity gains could be substantial but unevenly distributed within and across countries. - Advanced economies are better positioned to capture gains because of stronger preparedness and higher access. - Transition dynamics could strain fiscal frameworks through erosion of labor tax bases and rising social spending needs—particularly if employment-linked social insurance becomes less effective. - Macrofinancial vulnerabilities could rise as expectations and valuations adjust ahead of realized gains. ### Macroeconomic and financial implications (summary) - AI can accelerate research and development and automation, leading to potential large productivity gains contingent on diffusion and bottleneck resolution. - Distributional outcomes likely to be unequal across occupations and countries, with risks of increased inequality and market concentration. - Fiscal pressures could arise from reduced labor tax bases and higher social spending demands. - Financial-sector and valuation risks may emerge as markets price future productivity improvements before they materialize. ### Cross-cutting policy lessons (framework for preparedness) - Strengthen social spending and tax systems to manage distributional impacts and potential erosion of traditional labor tax bases. - Upgrade monetary diagnostics and scenario tools to operate under noisier price signals and uncertainty about the neutral rate. - Expand supervisory technology and macroprudential monitoring to address new transition risks stemming from AI-driven changes in the economy and financial sector. - Reinforce competition policy, support diffusion, and pursue international coordination on standards and taxation to mitigate concentration and cross-border spillovers. ### Introduction — core framing and evidence - Workshop combined technical frontier discussions (Day 1) and a closed-door scenario-planning exercise (Day 2) with around 50 participants, including AI technology experts, academics, and economists. - AI characterized as a macro‑critical transition; human decisions by managers, workers, regulators, and investors shape adoption sequencing, acceptance, and political sustainability. - Frontier capability timelines have compressed; estimates of 8–10 years for reaching artificial general intelligence are viewed as conservative by segments of the tech community (the “San Francisco Consensus”). - Access costs declining: inference prices for certain frontier models have dropped by over 99 percent. - Expertise (distinct from formal education) is central to wages; expertise accounts for roughly one‑third of observed wage variation. ### Growth, productivity, and debt dynamics - Channels through which AI influences growth: task automation, within‑task productivity gains, and accelerated R&D. - Aggregate effects depend on interactions among channels; low elasticities of substitution, bottlenecks, and diminishing returns can limit economy‑wide gains. - Binding constraints include energy demand, non‑automatable complementary tasks, institutional coordination, and capital accumulation limits. - Automation of AI research may produce strong feedbacks; within semi‑endogenous growth frameworks, outcomes can be bounded or explosive depending on feedback strength and returns to scale. - Expectations about AI‑driven growth can affect real interest rates and debt dynamics ahead of realized gains; current real interest rates provide limited evidence that markets are pricing in near‑term, extremely rapid AI‑driven growth. - AI can raise output via capital accumulation, but the share of tasks automated is likely more important for growth than productivity gains within individual tasks. ### Labor market impacts and adjustment - Labor outcomes depend on how AI alters task composition and expertise requirements, not only on job displacement. - Automation of high‑expertise tasks can raise employment but stagnate or reduce wages. - Automation of low‑expertise tasks can raise average expertise and wages for remaining workers but reduce total employment in that occupation. - Policy directions recommended: - facilitate labor reallocation (retraining, job matching, wage insurance); - public investment in complementary sectors; - strengthened social protection (transitional adjustment assistance, expanded wage insurance); - consider broader income support if automation diffuses widely. - Implementation challenges: low take‑up, administrative complexity, social stigma; reskilling may be unreliable if target occupations are shrinking. ### International inequality, market power, and spatial effects - Advanced economies better positioned to benefit from AI due to higher exposure, digital infrastructure, AI‑ready labor pools, and stronger institutions—risking widening cross‑country inequalities and “winner‑take‑most” dynamics. - Economies of scale in frontier models increase barriers to entry and market power; generative AI requires large compute and proprietary datasets concentrating power among a few firms and hyper‑scalers. - AI‑driven automation may erode traditional development pathways (reshoring to advanced economies), reallocating capital toward jurisdictions with stronger AI readiness, regulatory clarity, and reliable energy access. ### Society, governance, and policy architecture - AI governance frameworks lag rapid development: limited standards, reporting, and enforcement mechanisms complicate balancing innovation, safety, and democratic accountability. - Regulatory tools discussed: third‑party audits, transparency and disclosure requirements, licensing or liability regimes, with a shift toward pre‑deployment oversight. - Existing labor market and social protection tools are likely insufficient; need for human‑centric upgrades to unemployment insurance and transition support, and consideration of nonstandard policies (for example, universal basic income) under extreme scenarios. - Data gaps on AI deployment and labor impacts constrain real‑time policy calibration. - Cross‑border risks call for multilateral approaches in standards, surveillance, and policy sequencing. ### Scenario structure and selected scenario details - Shared assumption: advances in AI technologies yield capacity to perform a wide range of cognitive and manual tasks by the early 2030s. - Scenario 1: Baseline Diffusion - Society-wide adoption remains uneven, constrained by societal pushback, government interventions, and infrastructure bottlenecks. - Many countries adopt restrictive AI-related regulations that increase legal risks and compliance costs and enforce strong job protection programs. - The expansion of data centers continued throughout 2026; in 2027 financial market expectations changed as immediate financial returns from AI adoption underwhelmed, affecting AI-intensive firms financed with debt and equity. - Integration of AI into fiscal and monetary policy operations is guarded. - Scenario 2: Runaway Diffusion - AI diffuses rapidly across sectors under minimal regulatory constraints, propelled in part by intelligent AI agents. - AI technologies remain dominated by a small number of large firms that train better models, collect better data, and hold a disproportionate share of computation resources. - AI is deeply integrated into robotics and automates manufacturing, construction, mining, supply-chain management, and transportation. - AI technologies are adopted into fiscal and monetary policy operations, such as data analysis, forecasting, fiscal operations, and policy decisions. ### Macroeconomic and financial impacts under current policies - Near‑term (next five years): AI expected to boost productivity and output, with stronger effects under runaway diffusion. - Baseline scenario: growth effects are weaker and highly uneven across industries due to institutional frictions and social pushbacks. - Labor markets: rapid diffusion leads to broad job displacement; some high‑skilled workers benefit while capital owners capture large shares of gains; younger, less‑experienced workers face higher displacement risk and labor force participation may decline. - Financial stability risks highlighted: - household balance-sheet weakening and default risks for consumer credit exposures; - AI‑driven trading strategies heightening herding, decoupling markets from fundamentals, amplifying bubbles, and increasing volatility; - elevated leverage to finance AI investment and rapid capital obsolescence increasing uncertainty about future earnings and asset valuations; - concentration in the technology sector elevating single‑source exposures and undermining diversification. ### Policy implications and recommended directions - Fiscal policy: - protect tax capacity and fairness if labor income becomes less central; - manage pressures on public debt dynamics during the transition; - shift taxation toward capital and rent taxation while avoiding blunt “robot taxes”; - leverage AI for tax administration and enforcement; - de‑link safety nets from employment status and scale transfer systems quickly to target vulnerable groups; consider UBI‑type programs under extreme scenarios; - targeted public investment in energy, data centers, connectivity, and shared AI infrastructure to ease diffusion frictions and support SME adoption. - Monetary policy: - adapt to greater uncertainty around equilibrium interest rates and noisier inflation signals; - increase use of scenario‑based analysis and consider new tools under a digitalized economy and finance; - assess implications of digital money and central bank digital currency–type instruments for mandates and policy transmission. - Financial stability and supervision: - enhance visibility into AI‑related systemic risks; expand macroprudential buffers; - extend supervisory perimeters and crisis management tools to systemically important nonbanks while managing moral hazard via ex‑ante oversight and loss‑absorbing capacity; - adopt forward‑looking, scenario‑based stress testing and strengthen capital and liquidity buffers; - improve data collection on financial institutions’ exposures to AI‑sensitive sectors and their AI adoption. - Structural and international policy: - address concentration via competition policy and development of public‑good AI capabilities and regulatory bodies; - upgrade education and training systems for skill complementarities with AI; reform labor contracts to share gains and reduce resistance; - pursue international cooperation on standards, taxation principles, and interoperable assurance frameworks to avoid bifurcation and support emerging and developing economies. ### Main takeaways and operational priorities - No single baseline AI future; pervasive uncertainty in direction, speed, and diffusion requires robust scenario planning and policies that perform across plausible futures. - Outcomes depend as much on institutions, timing, and policy sequencing as on technological progress; without proactive institutional responses, gains likely concentrate and inequalities may worsen. - Transition dynamics can generate significant macrofinancial risks well before long‑run gains are realized—valuation cycles, leverage, concentration, and balance sheet vulnerabilities can propagate shocks across borders. - State capacity is a binding constraint: tax administration, social protection delivery, supervisory technology, and data governance determine countries’ ability to capture rents and cushion adjustments. - Key operational priorities: - close data gaps and strengthen diagnostics for AI diffusion, sectoral concentration, and labor market exposure; - regular scenario refreshes to align analysis and policy responses with the advancing frontier; - build flexible, forward‑looking fiscal, monetary, and financial frameworks resilient to a wide range of diffusion paths and transition risks. *Source: IMF Note — Global Economic and Financial Implications of AI (Executive Summary; Introduction; Conclusions and Policy Implications).* ### Executive Summary _________________________________________________________ 2 ### Executive Summary ### Context and purpose - Synthesizes insights from a high-level workshop and a scenario-planning exercise on the global economic and financial implications of artificial intelligence (AI), hosted by the IMF in collaboration with the Economics of Transformative AI Initiative (EconTAI) at the University of Virginia (December 10–11, 2025). - Framing: treat AI as a macro-critical transition rather than a standard technology shock because of its potential to restructure the global economy through task automation and accelerated research and development. - Emphasis: macroeconomic path shaped not only by frontier capability but by speed and breadth of diffusion and institutional/infrastructure readiness. ### Key near-term constraints (workshop consensus) - Physical and organizational bottlenecks can limit economy-wide productivity gains even with rapid frontier improvement: - energy and grid capacity - data-center infrastructure - persistence of tasks requiring physical presence - Labor market outcomes depend on changes in task composition and expertise content of work, not only on headline job displacement; this implies potentially divergent wage and employment effects across occupations. - Market structure and rents matter: scale economies in compute and data can amplify “winner-take-most” dynamics, raising the risk of greater within- and cross-country inequality. - Governance and social cohesion are critical constraints: institutions’ ability to acquire and maintain public trust determines the stability of the transition. ### Scenario-based assessment (five-year horizon) - Two alternative diffusion trajectories were explored assuming rapid technological capacity progress: - Baseline scenario: - Adoption is uneven and constrained by regulatory, infrastructure, and organizational frictions. - Runaway diffusion scenario: - Adoption accelerates broadly, with rapid expansion of automation across services and industry, and more front‑loaded labor displacement. - Common projection themes across scenarios: - Productivity gains could be substantial but unevenly distributed within and across countries. - Advanced economies are better positioned to capture gains because of stronger preparedness and higher access. - Transition dynamics could strain fiscal frameworks through erosion of labor tax bases and rising social spending needs—particularly if employment-linked social insurance becomes less effective. - Macrofinancial vulnerabilities could rise as expectations and valuations adjust ahead of realized gains. ### Macroeconomic and financial implications (summary) - AI can accelerate research and development and automation, leading to potential large productivity gains contingent on diffusion and bottleneck resolution. - Distributional outcomes likely to be unequal across occupations and countries, with risks of increased inequality and market concentration. - Fiscal pressures could arise from reduced labor tax bases and higher social spending demands. - Financial-sector and valuation risks may emerge as markets price future productivity improvements before they materialize. ### Cross-cutting policy lessons (framework for preparedness) - Strengthen social spending and tax systems to manage distributional impacts and potential erosion of traditional labor tax bases. - Upgrade monetary diagnostics and scenario tools to operate under noisier price signals and uncertainty about the neutral rate. - Expand supervisory technology and macroprudential monitoring to address new transition risks stemming from AI-driven changes in the economy and financial sector. - Reinforce competition policy, support diffusion, and pursue international coordination on standards and taxation to mitigate concentration and cross-border spillovers. *Source: IMF Note — Global Economic and Financial Implications of AI (Executive Summary).* ### Introduction ### Introduction ### Workshop purpose and scope - IMF convened the workshop “Global Economic and Financial Implications of Artificial Intelligence,” co‑organized with Economics of Transformative AI Initiative (EconTAI) and strategic foresight experts, to analyze evolving AI landscapes through technical presentations and structured scenario‑planning exercises. - The workshop combined technical frontier discussions (Day 1) and a closed‑door scenario‑planning exercise (Day 2) with around 50 participants, including AI technology experts, academics, and economists. - The Note summarizes key takeaways from the scenario exercise and lessons from the workshop, emphasizing the need to bridge high‑level analysis and actionable, real‑time operational guidance. ### Core framing and policy challenge - AI is characterized as a macro‑critical transition rather than a standard technology shock; human decisions by managers, workers, regulators, and investors shape adoption sequencing, acceptance, and political sustainability. - State capacity (tax systems, social protection, financial supervision, data governance) is a binding constraint on inclusive outcomes. - Central policy challenge: ensure macroeconomic and institutional frameworks are flexible, forward‑looking, and resilient across a wide range of diffusion paths and transition risks. ### Frontier capabilities and adoption frictions - Frontier capability timelines have compressed; estimates of 8–10 years for reaching artificial general intelligence are viewed as conservative by segments of the tech community (the “San Francisco Consensus”). - Transition from capability to economic impact is constrained primarily by institutional and organizational frictions: - Regulatory uncertainty, compliance burdens, organizational inertia, trust and behavioral barriers (model reliability concerns, managerial risk aversion). - Inadequate or excessively complex regulation can reinforce market concentration by favoring dominant firms able to absorb compliance costs. - Access costs declining: inference prices for certain frontier models have dropped by over 99 percent. - Adoption remains uneven across firms, sectors, and countries; economic gains likely concentrated among organizations with higher readiness. ### Growth, productivity, and debt dynamics - Channels through which AI influences growth: task automation, within‑task productivity gains, and accelerated R&D. - Aggregate effects depend on interactions among channels; low elasticities of substitution, bottlenecks, and diminishing returns can limit economy‑wide gains. - Binding constraints include energy demand, non‑automatable complementary tasks, institutional coordination, and capital accumulation limits. - Automation of AI research may produce strong feedbacks; within semi‑endogenous growth frameworks, outcomes can be bounded or explosive depending on feedback strength and returns to scale. - Expectations about AI‑driven growth can affect real interest rates and debt dynamics ahead of realized gains; current real interest rates provide limited evidence that markets are pricing in near‑term, extremely rapid AI‑driven growth. - AI can raise output via capital accumulation, but the share of tasks automated is likely more important for growth than productivity gains within individual tasks. ### Labor market impacts and adjustment - Labor outcomes depend on how AI alters task composition and expertise requirements, not only on job displacement. - Automation of high‑expertise tasks can raise employment but stagnate or reduce wages. - Automation of low‑expertise tasks can raise average expertise and wages for remaining workers but reduce total employment in that occupation. - Expertise (distinct from formal education) is central to wages; expertise accounts for roughly one‑third of observed wage variation. - Economy‑wide effects depend on substitution between intelligence‑intensive and physical sectors; outcomes hinge on degree of substitutability and physical constraints. - Policy direction: focus on facilitating labor reallocation (retraining, job matching, wage insurance), public investment in complementary sectors, strengthened social protection (transitional adjustment assistance, expanded wage insurance), and possibly broader income support if automation diffuses widely. - Implementation challenges: low take‑up, administrative complexity, social stigma; reskilling may be unreliable if target occupations are shrinking. ### International inequality, market power, and spatial effects - Advanced economies better positioned to benefit from AI due to higher exposure, digital infrastructure, AI‑ready labor pools, and stronger institutions—risking widening cross‑country inequalities and “winner‑take‑most” dynamics. - Economies of scale in frontier models increase barriers to entry and market power; generative AI requires large compute and proprietary datasets concentrating power among a few firms and hyper‑scalers. - AI‑driven automation may erode traditional development pathways (reshoring to advanced economies), reallocating capital toward jurisdictions with stronger AI readiness, regulatory clarity, and reliable energy access. - Without coordinated international intervention, the digital frontier may widen global inequality and create bifurcation risks. ### Society, governance, and policy architecture - AI governance frameworks lag rapid development: limited standards, reporting, and enforcement mechanisms complicate balancing innovation, safety, and democratic accountability. - Regulatory tools discussed: third‑party audits, transparency and disclosure requirements, licensing or liability regimes, with a shift toward pre‑deployment oversight. - Existing labor market and social protection tools are likely insufficient; need for human‑centric upgrades to unemployment insurance and transition support, and consideration of nonstandard policies (for example, universal basic income) under extreme scenarios. - Data gaps on AI deployment and labor impacts constrain real‑time policy calibration. - Cross‑border risks call for multilateral approaches in standards, surveillance, and policy sequencing. ### Scenario‑Based Assessment: structure and shared assumption - Day 2 exercise used two diffusion scenarios sharing a common assumption of rapid frontier capability progress: - Shared assumption: advances in AI technologies yield capacity to perform a wide range of cognitive and manual tasks by the early 2030s. - Baseline diffusion: gradual and uneven adoption constrained by regulation, societal resistance, infrastructure bottlenecks, and skills gaps. - Runaway diffusion: rapid and widespread adoption with limited friction, leading to economy‑wide automation, significant labor displacement, and high concentration of economic power among a small number of large AI firms. - Group discussions: first examined implications under current policies; second examined potential government interventions and policy responses. ### Macroeconomic and financial impacts under current policies - Near‑term (next five years): AI expected to boost productivity and output, with stronger effects under runaway diffusion. - Baseline scenario: growth effects are weaker and highly uneven across industries due to institutional frictions and social pushbacks. - Labor markets: rapid diffusion leads to broad job displacement; some high‑skilled workers benefit while capital owners capture large shares of gains; younger, less‑experienced workers face higher displacement risk and labor force participation may decline. - Market structure and spatial distribution: high data and compute requirements concentrate core AI service providers and reinforce metropolitan hub advantages, increasing regional disparities. - Price dynamics ambiguous: supply expansion from automation vs weaker demand from displaced households yields uncertain net price effects. - Cross‑country reallocations: runaway diffusion accelerates reshoring and capital flows toward jurisdictions with favorable tax regimes, regulatory clarity, reliable energy, and data access. - Migration likely declines as labor demand in advanced economies falls in runaway scenarios; baseline may sustain migration toward less‑automated locales. - Financial stability risks: - Large‑scale job losses weaken household balance sheets and raise default risks, pressuring banks with consumer credit exposure. - AI‑driven trading strategies could heighten herding, decouple markets from fundamentals, amplify bubbles, and increase volatility. - Elevated leverage to finance AI investment and rapid capital obsolescence increase uncertainty about future earnings and asset valuations. - Concentration in the technology sector elevates single‑source exposures and undermines diversification. ### Policy implications and recommended directions - Broad policy aim: capture AI’s productivity gains while managing transition risks across fiscal, monetary, financial stability, and structural/international domains. - Fiscal policy: - Erosion of labor tax bases as employment and wage income decline; shifting taxation toward capital and rent taxation needed but complicated by capital mobility and market concentration. - Avoid blunt “robot taxes”; strengthen capital income taxation at the individual level with safeguards against base erosion; leverage AI for tax administration and enforcement. - Social spending pressures rise; de‑link safety nets from employment status and scale transfer systems quickly to target vulnerable groups. Consider UBI‑type programs under extreme scenarios. - Targeted public investment in energy, data centers, connectivity, and shared AI infrastructure can ease diffusion frictions and support SME adoption. - Monetary policy: - Greater uncertainty in signal extraction, noisier inflation signals, and uncertain neutral interest rates complicate traditional monetary policy. - Increased value of scenario‑based analysis; consideration of new tools under a digitalized economy and finance. - Digital money and central bank digital currency–type instruments could reshape central bank operations but pose tradeoffs for mandates and policy transmission, especially for smaller open economies. - Financial stability and supervision: - Enhance visibility into accumulation of AI‑related systemic risks; expand macroprudential buffers; contain leverage and excessive risk‑taking among banks and systemically important nonbank financial firms. - Extend supervisory perimeters and crisis management tools to systemically important nonbanks while managing moral hazard via ex‑ante oversight and loss‑absorbing capacity. - Adopt forward‑looking, scenario‑based stress testing and strengthen capital and liquidity buffers given valuation and concentration uncertainties. - Improve data collection on financial institutions’ exposures to AI‑sensitive sectors and their AI adoption. - Structural and international policy: - Address concentration of AI service providers via competition policy and development of public‑good AI capabilities and regulatory bodies. - Upgrade education and training systems for skill complementarities with AI; reform labor contracts to share gains and reduce resistance. - International cooperation on standards, taxation principles, and interoperable assurance frameworks is pivotal to avoid bifurcation and to support emerging and developing economies. ### Main takeaways - No single baseline AI future; pervasive uncertainty in direction, speed, and diffusion requires robust scenario planning and policies that perform across plausible futures. - Outcomes depend as much on institutions, timing, and policy sequencing as on technological progress; without proactive institutional responses, gains likely concentrate and inequalities may worsen. - Transition dynamics can generate significant macrofinancial risks well before long‑run gains are realized—valuation cycles, leverage, concentration, and balance sheet vulnerabilities can propagate shocks across borders. - State capacity is a binding constraint: tax administration, social protection delivery, supervisory technology, and data governance determine countries’ ability to capture rents and cushion adjustments. - Rapid diffusion strains existing policy frameworks; macroeconomic and institutional frameworks must be flexible, forward‑looking, and resilient to manage transition risks rather than predict a single AI future. *Source: insea2026002 - Introduction* ### Conclusions and Policy Implications ### Conclusions and Policy Implications ### Core findings on AI transition dynamics - AI offers large potential gains, but the transition will test fiscal and monetary frameworks, financial stability, and social cohesion. - The workshop’s core message: diffusion speed and complementarities—not frontier capability alone—will shape macrofinancial outcomes. - The immediate policy challenge is managing adjustment dynamics under deep uncertainty rather than forecasting a single endpoint. - These dynamics are fundamentally shaped by a sequence of human decisions whose expectations, incentives, and perceptions of legitimacy dictate the adoption sequence. ### Fiscal policy implications - Fiscal policy should: - protect tax capacity and fairness if labor income becomes less central; - manage pressures on public debt dynamics during the transition. - Fiscal analysis priority: incorporate scenarios with persistent declines in labor income share. ### Monetary policy implications - Monetary policy frameworks need to adapt to: - greater uncertainty around equilibrium interest rates; - noisier inflation signals driven by relative price adjustment. - Monetary and financial policy frameworks could account for weaker transmission and new sources of volatility. - Integration of AI into fiscal and monetary policy operations is mentioned as guarded in Scenario 1 and actively adopted in Scenario 2. ### Financial stability and macroprudential oversight - Macroprudential oversight should extend to third party and vendors as risk migrates across institutions and infrastructures. - Regulatory frameworks must avoid inadvertently reinforcing market concentration through overly complex or poorly calibrated rules. - Safety and reporting standards should promote interoperability while limiting fragmentation and contagion. ### Industrial policy and infrastructure - Industrial policy has a role in addressing market deficiencies in scaling up clean power and smart electric grids with rising compute power demands, especially where the private sector cannot provide them alone. ### Behavioral frictions and adoption dynamics - Behavioral frictions (job loss fears, compliance liability, reputational harm, risk aversion, doubts about model reliability) can: - make adoption hesitant and uneven; - delay productivity gains; - prolong structural instability and periods of financial stress. - Adoption depends on how workers and firms respond to uncertainty and perceived risks. ### Social cohesion, implementation, and trust - Inclusive outcomes require policy designs that explicitly account for implementation and trust. - Governments need adjustment mechanisms that build and sustain social consensus, particularly when distributional impacts are front-loaded and concentrated. - Clear and transparent public communication and expectations management should support policy buy-in and public trust. - Policies that ease labor market transitions and support skill adaptation are more likely to attenuate political and social pressures than delayed support or weak delivery. - Effective design must account for implementation challenges—including low take-up, administrative complexity, and social stigma—that can undermine policy effectiveness. ### Data, measurement, and diagnostics - A key priority: close data gaps and strengthen diagnostic capabilities to support timely policy responses during the AI transition. - Workshop participants emphasized systematically tracking indicators of: - AI diffusion; - sectoral concentration; - labor market exposure across countries in near real time. - Critical data gaps include AI-related investment, pricing, occupational exposure, adoption, and usage. - Improved measurement of AI adoption and usage, combined with better data on task-level impacts and investment flows, would help distinguish between gradual and more rapid diffusion paths and support earlier identification of emerging macrofinancial risks. ### Scenario planning takeaways - The scenario-planning exercise explored a range of plausible diffusion paths under accelerated timelines where AI attains human expert-level capabilities over the next five years and by the early 2030s can perform most cognitive and physical tasks. - About 30 IMF staff and 15 external experts participated; they were split into 6 groups of 7–8 participants. - Scenario 1: Baseline Diffusion - Society-wide adoption remains uneven, constrained by societal pushback, government interventions, and infrastructure bottlenecks. - Many countries adopt restrictive AI-related regulations that increase legal risks and compliance costs and enforce strong job protection programs. - AI diffusion in services is significant but not transformative due to occupational and licensing policies and costly human supervision requirements. - Robot diffusion in mining, construction, and manufacturing is hindered by strict regulations such as safety standards and tight job protection requirements. - The expansion of data centers continued throughout 2026; in 2027 financial market expectations changed as immediate financial returns from AI adoption underwhelmed, affecting AI-intensive firms financed with debt and equity. - Integration of AI into fiscal and monetary policy operations is guarded. - Scenario 2: Runaway Diffusion - AI diffuses rapidly across sectors under minimal regulatory constraints, propelled in part by intelligent AI agents. - AI technologies remain dominated by a small number of large firms that train better models, collect better data, and hold a disproportionate share of computation resources. - The private sector drives diffusion focused on cost optimization and profit maximization; in many parts of the economy humans become optional rather than necessary. - AI is deeply integrated into robotics and automates manufacturing, construction, mining, supply-chain management, and transportation. - AI technologies are adopted into fiscal and monetary policy operations, such as data analysis, forecasting, fiscal operations, and policy decisions. ### Multilateral and forward-looking priorities - Multilateral surveillance should consider AI as a potentially asymmetric global shock with implications for spillovers and policy coordination. - Priorities underscore the benefit of flexible, forward-looking frameworks resilient to a wide range of diffusion paths and transition risks, grounded in evolving macrofinancial reality. - Regular scenario refreshes are recommended to keep analysis and policy responses aligned with the advancing frontier. *IMF Note — Global Economic and Financial Implications of AI: Conclusions and Policy Implications.* --- _Source: https://www.imf.org/-/media/files/publications/imf-notes/2026/english/insea2026002.pdf_