Introduction
It is a pleasure to join you at the Financial Street Forum. Thank you for the invitation.
Last year at this forum, I discussed how new technologies, such as tokenisation, can help to reform the governance of the financial system’s core infrastructure while upholding the principles that ensure money remains stable and trustworthy. This year, as new technologies have played an increasingly important role, I would like to continue that discussion by focusing on artificial intelligence – or AI – and its implications for central banking.
I will consider the following three implications of AI: for the macroeconomy and financial stability; for central banks and financial regulators; and for international cooperation, including the role of international financial organisations such as the Bank for International Settlements (BIS).
Implications for the macroeconomy and financial stability
Let me start with implications for the macroeconomy and financial stability. There are many, but I will focus on three: (i) how AI may affect price formation; (ii) how it may contribute to herding; and (iii) how it may change cyber risk. All of these are core issues for central banks and regulators.
First, price formation. AI gives businesses new ways to process information and adjust prices. As AI tools are adopted more widely, firms may respond more frequently to changes in costs, demand or market conditions.
There is some evidence of this at the firm level. Online retailers using algorithmic pricing can change prices more frequently and set more uniform prices across locations. They can respond quickly to changes in factors such as gas prices or exchange rates.
The macroeconomic significance is that aggregate shocks may feed through to consumer prices more quickly, potentially changing inflation dynamics.
Second, herding, with implications for financial stability.
Common expectations can support investment and innovation, but they can also concentrate exposures. Financial institutions may use similar models, data or decision rules in trading, lending and risk management. If those tools lead them to interpret information similarly, they may take similar actions at the same time, creating crowded trades or synchronised changes in lending and risk appetite.
Such correlated responses could reinforce one another. If institutions reduce risk or sell similar assets at the same time, price movements could be amplified. Leverage and assets that are difficult to sell could intensify the adjustment. This is a potential channel through which AI use could contribute to correlated behaviour across the wider financial system.
Indeed, a recent BIS analysis points to a growing role for debt, including private credit, and has identified circular financing among firms within the AI ecosystem, including commercial links between firms and suppliers or the customers they invest in.
Third, cyber risk. AI can strengthen cyber defences, but it can also make attacks faster, more powerful and more complex. It can help malicious actors create convincing messages, automate parts of an attack or identify vulnerabilities more quickly. This may leave institutions with less time to detect potential attacks and respond.
The consequences can extend beyond the institution directly targeted. Financial firms rely on shared technology, data and service providers. A disruption affecting a critical provider could therefore affect several institutions at once. The concern is not simply that an individual attack may become more sophisticated; it is that its speed, reach and potential impact may increase in a highly connected financial system.
Implications for central banks and regulators
These are important challenges that central banks and financial regulators must increasingly contend with. They affect how these institutions assess economic conditions, monitor financial risks and safeguard resilience.
So how should central banks and regulators respond?
On price formation, more frequent price changes increase the value of timely and detailed information. Central banks can use new data and analytical methods to monitor developments in prices and to understand how firms are responding to changing conditions. AI can help organise large amounts of information and identify patterns that might otherwise be difficult to see. This can strengthen analysis, but it does not determine what those patterns mean for inflation or what policy response is appropriate. Economic interpretation and human judgment remain essential.
On herding, the implication is to look beyond individual institutions and assess whether exposures or dependencies are becoming common across the financial system. Authorities can monitor AI-related investment and financing, as well as institutions’ reliance on similar models, data and providers. The BIS’s 2026 Annual Economic Report calls for better supervisory visibility into AI use, including to identify model-driven correlated exposures and herding.
This also has a macroprudential dimension. Authorities could examine how the system might respond to a sharp repricing of AI-related assets, a disruption at an important provider or similar model-driven decisions across institutions during stress. Such analysis can help reveal where exposures are shared and how pressures might spread. The point is to strengthen visibility and preparedness; it is not to imply that one particular macroprudential measure is required in every jurisdiction.
On cyber risk, AI can give central banks, supervisors and financial institutions better tools to identify threats and support a faster response. At the same time, institutions need robust cyber risk management and operational resilience. This includes clear responsibilities, timely decisions, effective patching and the ability to respond to and recover from incidents. Authorities also need to understand institutions’ reliance on critical third-party providers. The Financial Stability Institute’s recent work points to strengthening existing cyber resilience frameworks as AI changes the threat environment.
Implications for international cooperation and the BIS
Finally, let me turn to international cooperation and the role of international organisations, including the BIS.
The effects of AI will not stop at national borders. Financial markets are closely connected, technology providers operate internationally, and cyber incidents can spread rapidly across institutions and jurisdictions. National responses alone will therefore not be sufficient.
At the same time, central banks and regulators are at different stages in their use and supervision of AI. This is particularly true for governance and human judgment. AI can make analysis more efficient, but its outputs can be wrong, difficult to explain or shaped by limitations in the data. Authorities are still developing approaches to ensure genuine human scrutiny, clear accountability and independent judgment. By exchanging experience, they can learn what works, identify common problems and avoid repeating one another’s mistakes.
There is also an important capacity-building dimension. Not every authority has the same access to data, technology or specialised expertise. International cooperation can help to narrow these gaps through training, technical assistance and the sharing of knowledge and practical experience.
This is where the BIS can contribute. Through its forums, networks and BIS Innovation Hub Centres, the BIS brings central banks and regulators together to compare approaches to AI governance, human oversight and risk management. It can also support capacity-building by pooling expertise and disseminating lessons across jurisdictions.
One example is Project Spectrum, a BIS Innovation Hub collaboration with the Deutsche Bundesbank and the European Central Bank. It explored how AI could help turn large volumes of daily product-price data into structured information that can support inflation analysis. Projects such as this allow central banks to experiment together, learn from practical experience and share the resulting knowledge more widely.
Conclusion
Let me conclude. AI may change how quickly prices respond to information; it may contribute to correlated behaviour; and it may increase the speed and reach of cyber threats. These developments matter for central banks and regulators, both in their analysis of the economy and in their oversight of the financial system.
The task is to use AI where it can improve information and analysis, while maintaining sound judgment, governance and resilience.
International cooperation can help authorities develop sound governance, preserve human judgment and build the capacity needed to use AI responsibly. The BIS will continue to support this collective effort through dialogue, research, practical experimentation and capacity-building.
Thank you.
