The International Monetary Fund (IMF) has urged central banks and financial regulators in the global financial ecosystem to strengthen their oversight of artificial intelligence (AI), since the rapid adoption of the technology across the financial sector could amplify systemic risks if allowed without adequate governance.
The Bretton Woods financial institution, in a blog published shared via its official X account, noted that AI had been rapidly transforming how financial institutions price risk, allocate credit, conduct trading and supervise markets, making it increasingly central to the global financial system.
The Fund maintained that while AI had the potential to improve efficiency, liquidity, fraud detection and market surveillance, regulators must move quickly to ensure the technology strengthens, rather than undermines, financial stability.
The blog, authored by Tobias Adrian, Director of the IMF’s Monetary and Capital Markets Department, listed three immediate priorities for central banks and financial supervisors to adopt in order to prevent their financial systems to the AI risks.
- Advertisement -
Specifically, the IMF listed the priorities as including strengthening oversight and governance of AI-driven trading, lending and supervisory technology; improving visibility into AI use and the risks posed by shared models and synchronized trading strategies; and deepening international cooperation on cyber resilience and operational security.
The Fund maintained that AI was compressing the speed at which financial markets operate, with trading decisions, lending assessments and supervisory analytics increasingly taking place in real time.
Noting that while this can improve market efficiency under normal conditions, the IMF stressed that it could also allow financial shocks to spread more rapidly during periods of stress.
The Fund noted that AI-powered investment strategies could amplify market volatility if multiple systems react simultaneously to similar market signals, warning that future market disruptions could stem less from programming errors and more from large numbers of AI models responding in parallel to the same information.
The IMF also highlighted growing concerns over the opacity of some AI models, saying even sophisticated financial institutions may struggle to explain why certain AI-driven decisions were made during market stress, pointing out that this could make it more difficult for regulators to detect emerging vulnerabilities and respond effectively.
Beyond trading and lending, the IMF clarified that AI was becoming deeply integrated into banking operations, payment systems, exchanges and clearing houses and cautioned the central banks and financial system regulators that heavy reliance on a small number of cloud computing, data and AI model providers could create significant concentration risks, where disruptions at a single provider could affect multiple financial institutions simultaneously.
In addition, the multilateral development finance institution further expressed concerns that generative AI was accelerating the sophistication of cyber threats, making phishing attacks, fraud schemes and cyber intrusions more difficult to detect and contain, and urged central banks to strengthen cyber resilience frameworks, conduct AI-focused stress exercises and invest in defensive AI technologies.
On financial supervision, the IMF noted that several central banks, including those in France, Germany, Portugal, Japan, Canada and the United States, were already deploying AI-powered supervisory technology to improve market surveillance and identify emerging risks.
However, it stressed that AI should support—not replace—human judgment in regulatory decision-making.
In its concluding remarks, the IMF projected that financial stability in an AI-driven world would depend not only on the technology itself but also on the institutions, governance frameworks and safeguards guiding its use, calling for early and coordinated policy action to prevent future financial instability from becoming faster, more interconnected and harder to manage.
