The International Monetary Fund has placed safeguards at the center of its artificial-intelligence policy agenda as rapid adoption changes investment, labor markets, financial systems, and the distribution of income. At an October 16 Annual Meetings seminar, the Fund framed the policy challenge around capturing productivity gains while containing economic, financial, and social risks.
The official IMF event page identified technological disruption, market concentration, cyber risk, and financial-stability risk as central concerns. It also called attention to worker skills, economic resilience, and policies that spread the gains from AI more widely. IMF Managing Director Kristalina Georgieva joined government, central-bank, business, and technology leaders for the discussion.
The Fund’s position combines two conclusions. AI can raise productivity and expand output, but the benefits will depend on infrastructure, skills, competitive markets, and credible governance. Countries that enter the transition with weak power systems, limited digital access, narrow fiscal space, or fragile labor-market institutions face a harder adjustment.
A large investment cycle with uncertain returns
The scale of the current AI investment cycle gives the debate immediate macroeconomic relevance. The IMF’s 2026 Annual Report estimates that technology investment related to AI added about 0.5 percentage point to US economic growth in 2025. It also cites external estimates suggesting private-sector AI investment could exceed $2 trillion worldwide in 2026.
That spending supports demand for data centers, advanced semiconductors, power equipment, cooling systems, construction, and specialized labor. It can also improve productivity when firms redesign processes around the technology. The near-term boost therefore reaches well beyond software companies.
Large capital programs still need to earn adequate returns. The IMF warns that increasingly debt-financed AI investment could disappoint if expected productivity or revenue fails to arrive. Sharp valuation declines could weaken household wealth, cut corporate spending, and trigger layoffs. Circular financing within the AI supply chain adds another vulnerability when a small group of companies act as customers, investors, and financiers for one another.
For lenders, the relevant question is how much debt service depends on aggressive demand forecasts, high utilization, and continued access to cheap funding. Data centers and power assets have long lives, while technology demand can change quickly. Projects with contracted revenue, diversified customers, and realistic electricity costs should withstand a repricing better than capacity built mainly on optimistic growth assumptions.
Productivity gains will arrive unevenly
Georgieva has previously set out the size of the opportunity. In February, she said AI could lift global productivity growth by as much as 0.8 percentage point a year. Her official IMF remarks also estimated that AI could affect 40 percent of jobs worldwide and 60 percent in advanced economies.
Exposure does not mean every affected job disappears. AI may automate tasks, change the skills required for a role, or create new work around cheaper and more scalable services. IMF First Deputy Managing Director Dan Katz noted that the aggregate outcome depends on how economies adjust. His June analysis argues that lower costs can increase demand and output, while weak labor mobility and barriers to new firms can slow the creation of replacement jobs.
Distribution matters alongside the aggregate result. Workers with scarce AI skills may earn more, and some lower-skill service roles may benefit from spending by higher earners. Middle-skill and entry-level jobs can come under pressure if firms automate routine analytical and administrative tasks. Regions with fewer digital firms, weak connectivity, or expensive electricity may attract less investment and receive a smaller share of the productivity dividend.
These differences can affect public finances. Governments may face calls for retraining, income support, education reform, power investment, and digital infrastructure at the same time that automation changes payroll-tax receipts. Countries with high debt and limited borrowing capacity will have to sequence those investments carefully. Poorly targeted subsidies could add fiscal risk without improving national AI readiness.
Safeguards need to preserve useful innovation
The IMF’s AI policy overview calls for advanced economies to balance innovation with strong safeguards. It advises emerging and developing economies to expand digital infrastructure and skills while using smarter regulation, public investment, and international coordination to narrow the resilience gap.
That approach favors rules tied to identifiable risks. Cybersecurity, consumer protection, data governance, competition, and the use of AI in regulated financial activities require clear responsibilities. Firms need to know who approves a model, monitors its performance, handles incidents, and remains accountable when automated decisions cause harm.
Regulation can also shape market structure. Compliance costs that only the largest technology companies can absorb may reinforce concentration. Weak rules create a different problem by allowing unsafe systems or opaque dependencies to spread through critical services. Authorities need standards that can be applied by smaller firms, tested in practice, and updated as capabilities change.
International coordination is especially important because AI models, cloud platforms, payment systems, and cyber threats cross borders. A failure at one widely used provider can affect many institutions and countries at once. Common terminology, incident reporting, testing practices, and recovery plans can reduce confusion during a disruption.
Financial systems face faster cyber and market shocks
AI changes the speed and scale of financial risk. Machine-learning tools can improve fraud detection, credit analysis, trading, and supervision. The same technology can accelerate vulnerability discovery, phishing, malware development, and attack coordination.
An IMF note on AI and financial-sector cybersecurity identifies shared digital infrastructure and common service providers as important transmission channels. AI-enabled attacks can spread more quickly across similar systems, leaving institutions less time to identify, contain, and recover from a breach. The note recommends governance and technical controls that limit the scope of damage, stronger recovery capacity, and deeper international cooperation.
Financial markets face a related coordination problem. Models trained on similar data and optimized for similar signals may react in the same direction during stress. Automated execution can then amplify price moves and liquidity gaps. Supervisors need visibility into model use, external dependencies, and the concentration of exposures across institutions.
Banks and nonbank lenders should treat AI infrastructure and technology-company credit as a connected portfolio risk. Borrowers may share semiconductor suppliers, cloud providers, power grids, and investor funding. A disruption in one layer can weaken several companies at once even when their immediate business models appear different.
Policy choices will differ by country
Advanced economies generally have deeper capital markets, stronger digital infrastructure, and more fiscal capacity to support research and worker adjustment. They also hold larger exposures to highly valued technology companies and complex financial markets. Their policy mix should emphasize competition, operational resilience, model governance, and labor mobility.
Many emerging economies have a different opportunity. AI can improve tax administration, customs, health services, education, and access to finance without requiring every country to build a frontier model. Benefits will depend on affordable electricity, connectivity, secure digital identity, usable data, and the capacity of public institutions to buy and supervise technology.
Import dependence creates another source of risk. Countries that rely on foreign cloud services, chips, models, and technical expertise can face higher costs or interrupted access during geopolitical disputes. Diversification and interoperability may improve resilience, although duplicating infrastructure can be expensive.
Analyst’s View
First, credit analysis should separate AI demand from borrower quality. Strong sector growth can coexist with weak individual projects. Lenders should test utilization, pricing, electricity costs, refinancing needs, customer concentration, and the value of collateral under a technology downturn.
Second, sovereign risk will depend on how governments finance the transition. Public investment in power, networks, education, and digital services can raise productive capacity. Debt-funded projects with unclear demand or large guarantees can weaken fiscal resilience. Transparent procurement and contingent-liability reporting will help investors distinguish capacity building from politically favored spending.
Third, market positioning should account for concentration. AI gains may accrue to a narrow group of firms and countries before spreading more widely. That can support earnings and asset prices for a period while increasing sensitivity to disappointing revenue, regulation, power constraints, or cyber incidents.
The IMF’s policy message is therefore practical: pursue the productivity opportunity, prepare workers and institutions for adjustment, and build safeguards before financial and digital dependencies become too large to manage safely.
