Central banks expect artificial intelligence to have its greatest impact on cybersecurity, while fewer anticipate a major role in monetary policy decisions. A 2026 Bank for International Settlements (BIS) survey found that about 36% of respondents expect AI to have a large impact on cybersecurity over the next two years, compared with roughly 14% for monetary policy.
The October report, AI Readiness in Central Banks, was prepared by the Irving Fisher Committee on Central Bank Statistics at the BIS. The survey covered institutions from 65 jurisdictions and organisations.
Cybersecurity tops expectations
Central banks operate sensitive financial infrastructure and handle confidential data. AI systems can analyse network traffic, detect anomalous behaviour and help security teams prioritise alerts. The technology may improve detection, but also creates risks as attackers adopt AI and organisations expose internal data to new models.
Economic research and statistics also rank highly among expected applications. Financial stability, payment systems and monetary policy are viewed more cautiously, reflecting the higher stakes and governance requirements of core central-bank operations.
AI can support, but not make, interest-rate decisions
AI can help economists process corporate announcements, prices, market information and economic reports. It may also support forecasting and scenario analysis. However, decisions about interest rates require assessments of the causes of inflation, economic conditions and risks that models cannot reliably resolve on their own.
In Poland, statutory responsibility for setting interest rates rests with the Monetary Policy Council. AI tools can inform its analysis but do not replace its authority or accountability.
Most reported projects are already operational
Statistics and economic research account for the largest numbers of AI projects reported by central banks. Communication-related projects now outnumber those in monetary policy, financial stability and cybersecurity.
More than two-thirds of reported AI projects in the 2026 survey had reached operational use, including limited-scale deployments, compared with around half in 2024. Yet respondents have become more measured about the technology’s potential impact as practical experience has exposed integration, security and verification challenges.
Costs and governance remain obstacles
Nearly 80% of respondents identified IT expenditure as an AI-related risk. Most central banks currently allocate less than 5% of their total budgets to AI, and actual spending has been lower than institutions expected two years earlier.
Only 15% of respondents rated their ability to monitor compliance with internal AI policies as high. Another 54% described it as moderate, 27% were developing appropriate controls and 4% reported limited capabilities.
The BIS highlights the need for auditability, clear access restrictions and human oversight, particularly as AI agents gain the ability to perform tasks across connected systems. For central banks, the ability to reconstruct what an automated system did and who authorised it is essential to operational resilience.
Source: BIS Irving Fisher Committee, AI Readiness in Central Banks, October 2026.