将人工智能用于政策目的

The use of artificialintelligence for policypurposes Report submitted to the G20 Finance Ministers andCentral Bank Governors October 2025 This publication is available on the BIS website (www.bis.org). ©Bank for International Settlements 2025. All rights reserved. Brief excerpts may be reproduced ortranslated provided the source is stated. The use of artificial intelligence for policy purposes BIS report to be submitted ahead of the October 2025 G20 FMCBGmeeting Contents 1.Introduction............................................................................................................................................................................ 22.Core concepts of AI............................................................................................................................................................. 33.AI use cases by central banks and supervisors......................................................................................................... 43.1 Information collection ............................................................................................................................................... 53.2 Macroeconomic and financial analysis to support monetary policy........................................................ 63.3 Oversight of payment systems................................................................................................................................ 73.4 Supervision and financial stability........................................................................................................................104.Challenges and lessons learned ...................................................................................................................................155.Conclusions ..........................................................................................................................................................................206.References.............................................................................................................................................................................22 1.Introduction The rapid adoption of artificial intelligence (AI) – broadly defined as computer systems capable of tasksthat normally require human intelligence – is poised to have a profound effect on the financial system andthe real economy (BIS (2024); Aldasoro, Doerr, Gambacorta and Rees (2024); Aldasoro, Gambacorta,Korinek, Shreeti and Stein (2024); IMF (2024)). The adoption of generative AI (gen AI), ie tools that engagewith text-based content and allow users to converse with computers through ordinary language, isproceeding at a speed that easily outpaces previous waves of technology adoption. ChatGPT alonereached one million users in less than a week and many firms are already integrating gen AI into their dailyoperations. To do so, they are investing heavily in AI technology to tailor it to their specific needs, and inmany cases they have embarked on a hiring spree of workers with AI-related skills. These developments,and the attendant effects on inflation, productivity, consumption, investment and labour markets are ofparamount concern to central banks and other supervisory and regulatory authorities (Aldasoro, Doerr,Gambacorta, Gelos and Rees (2024); Cazzaniga et al (2024)). Central banks were early adopters of machine learning (ML) (a key component of AI techniques),using it to gain insights for statistics, research and policy long before AI became a popular topic (Araujoet al (2024)).1Discussions on AI and ML at central banks are pervasive (Graph 1.A) and expected budgetallocations bear that interest out (Graph 1.B). Indeed, central banks, financial regulators and supervisoryauthorities regularly handle vast data sets and complex decision processes in pursuit of safeguardingmonetary and financial stability, and the integrity of payment systems. Today, the greater capabilities ofnew AI methods – such as the large language models (LLMs) underpinning gen AI – open furtheropportunities, from improved economic analysis to better regulatory oversight, potentially enhancing theeffectiveness and efficiency of these institutions and of policymaking more broadly. Source: IFC (2025). 1Machine learning is a term referring to techniques designed to detect patterns in data and use them in prediction or to aiddecision-making. This report examines how central banks and other supervisory institutions are leveraging AI forpolicy purposes. The report first offers a brief discussion of core AI concepts relevant to public authorityuse cases, focusing in particular on ML. It then provides examples of how central banks and supervisoryauthorities are already using big data and ML in four key areas. These are: (i) information collection andthe compilation of official statistics; (ii) macroeconomic and financial analysis in support of monetarypolicy; (iii) oversight of payment systems; and