《OMFIF：2025智能金融系統風險：AI對金融業的影響與挑戰研究報告（英文版）（14頁）.pdf》由會員分享，可在線閱讀，更多相關《OMFIF：2025智能金融系統風險：AI對金融業的影響與挑戰研究報告（英文版）（14頁）.pdf（14頁珍藏版）》請在三個皮匠報告上搜索。

1、Digital Monetary InstituteWHAT AI MEANS FOR FINANCERISKS OF AN INTELLIGENT FINANCIAL SYSTEMomfif.org23Executive summaryUnder-studied and under-examined4Chapter 1First-order risks7Chapter 2Second-order risks9Chapter 3Policy responses and future research11References12Further readingContentsOfficial Mo

2、netary and Financial Institutions Forum 6-9 Snow Hill,London,EC1A 2AYT:+44(0)20 700 27898enquiriesomfif.org omfif.orgABOUT OMFIFWith a presence in London,Washington and New York,OMFIF is an independent forum for central banking,economic policy and public investment a neutral platform for best practi

3、ce in worldwide public-private sector exchanges.AUTHORSJulian JacobsDoctoral Candidate,University of OxfordAndrew SuttonTrustee,London Initiative for Safe AI and Commercial BankerEDITORIAL AND PRODUCTIONSimon HadleyDirector,ProductionWilliam Coningsby-BrownProduction ManagerSarah MoloneyChief Subedi

4、torJanan JamaSubeditor 2025 OMFIF Limited.All rights reserved.Strictly no photocopying is permitted.It is illegal to reproduce,store in a central retrieval system or transmit,electronically or otherwise,any of the content of this publication without the prior consent of the publisher.While every car

5、e is taken to provide accurate information,the publisher cannot accept liability for any errors or omissions.No responsibility will be accepted for any loss occurred by any individual due to acting or not acting as a result of any content in this publication.On any specific matter reference should b

6、e made to an appropriate adviser.Company number:7032533.ISSN:2398-42363omfif.orgExecutive summaryTHE RAPID advancement of artificial intelligence may have significant impacts on the global economy.Leaps in machine learning development,with their potential for far-reaching impacts across multiple sec

7、tors,often evoke comparisons to the transformative power of the Industrial Revolution1.While the full impact of AI remains deeply uncertain,its growing significance as a general-purpose technology suggests that its effects on finance a sector deeply intertwined with economic activity may be profound

8、.Many scholars and policy-makers agree that AI could help improve efficiency in the financial system.The technology could broaden access to finance,making high-quality products more accessible and personalised,while also improving the quality of risk analysis2.Alongside the potential benefits,howeve

9、r,AI presents a spectrum of risks.A critical yet under-examined dimension of these risks lies in the interplay between AI and the financial system.The modern financial system is complex:it is both susceptible to fragility and capable of rapidly transmitting shocks across markets and borders.Regulato

10、ry frameworks that have evolved over decades to ensure financial stability may prove ineffective or even counterproductive in the face of swift AI-driven transformations.Such risks may arise both from the application of AI within the financial system itself or from how financial institutions and mar

11、kets respond to its application elsewhere.The use of digital tools to make decisions in the financial sector has long been widespread.Since the 1980s,increasingly sophisticated machine learning models have been used for financial predictions,portfolio management,algorithmic trading,credit decision-m

12、aking and other tasks3.Yet advances in AI,brought to public attention by the release of OpenAIs ChatGPT model in late 2022,have significantly increased the scope and degree of possible computer automation for many cognitive tasks.Advanced AI techniques are being used or tested for a wide range of fi

13、nance use cases.In banking,this includes anti-money laundering and fraud screening,credit and risk modelling,customer services and back-office automation.In insurance,AI is being used for underwriting,risk assessment and claims management.And in asset management,AI is being deployed for market signa

14、l analysis,risk modelling and automating trades.There is also discussion about the prospect of increasingly autonomous or semi-autonomous AI agents taking on trading decisions in financial markets,yet the exact extent of this remains uncertain4.There is little question that finance is vital to the m

15、odern economy and its effective functioning.Increasing the efficiency of capital allocation,facilitating transactions,supporting saving and investments,transferring and mitigating risks,generating helpful price signals and unlocking beneficial trade all contribute to spurring growth.Yet finance is a

16、lso prone to crisis.Risks can accumulate and spill over elsewhere due to the misallocation of capital and excessive risk-taking,causing greater vulnerability for many people.Yet,to date,AI risks in the financial sector have been under-studied.This paper attempts to synthesise existing research on AI

17、 risks within and through the finance system.While acknowledging the limitations of literature reviews particularly in a field with a limited selection of published peer-reviewed works we hope it will be a valuable foundation for future researchers interested in the topic of AI in finance.UNDER-STUD

18、IED AND UNDER-EXAMINEDIn finance,digital tools have been used for decades to improve decision-making.While the advent of artificial intelligence and its associated risks in the financial sector have been anticipated,they have not yet been fully explored.Advanced AI techniques are being used or teste

19、d for a wide range of finance use cases.omfif.org4AI POSES a range of first-order risks,which include hazards that might arise from the deployment of AI within the financial system and by bad actors using AI adversarially.Scams and impersonationsAI increases fraud risk,including by lowering the cost

20、 of running phishing-style attacks,which are increasingly hyper-personalised and effective scams.Deepfake technology can convincingly mimic individuals,making it hard to distinguish artificial voices or images from real ones.Researchers have demonstrated that AI can defeat voice password systems tha

21、t were once considered secure5,which raises major policy implications for consumer protection,liability and competition.Despite measurement difficulties,some datasets show there has been a fivefold increase in the number of deepfake fraud attempts from 2023 to 2024 globally6.Chapter 1FIRST-ORDER RIS

22、KSFrom operational challenges to market manipulation,incorporating artificial intelligence into the financial system introduces a slew of first-order risks.TypeRiskDescriptionBad actorsScams and impersonationsGenerative AI techniques could facilitate various types of fraud while undermining some exi

23、sting controls,such as voice passwords.It also may create new hazards,such as deepfake voice and video technology.Cybersecurity threatsAI-powered cybersecurity attacks may lead to an arms race between protective AI-driven measures that aim to stamp out cyberattacks and malicious actors leveraging AI

24、 to conduct cyberattacks.Market manipulationAI could be a powerful tool for disinformation,potentially serving to mislead investors or consumers.Data and model featuresPrivacy concernsModels used by financial institutions may be trained on datasets that include sensitive financial and non-financial

25、data,increasing the need for privacy protections.Lack of explainabilityMany AI models are opaque,making their decision-making processes difficult to understand.This can pose challenges for regulators and institutions that need to understand the rationale upon which sensitive decisions are taken.AI b

26、iasesAI datasets or algorithms can be biased,producing skewed outcomes that inaccurately disadvantage certain groups.Market stabilityAlgorithmic convergence and market instabilityMarket participants relying on similar AI models in asset allocation and portfolio management may increase trading correl

27、ations and volatility,and lead to herding behaviour,raising the risk of destructive events like flash crashes.Operational and regulatory challengesConcentration and third-party relianceReliance on a small number of firms and key geographic locations in the AI supply chain could create concentrated o

28、perational risks for financial institutions,with implications for wider system stability.Regulatory challengesRapid technological advances may outpace regulators ability to respond.Regulatory approaches may diverge,increasing costs and uncertainty for financial institutions,and widening opportunitie

29、s for regulatory arbitrage.Figure 1.First-order risks of AI in financeSource:OMFIF5omfif.orgCybersecurity threats and the arms raceClosely related are cybersecurity threats AI gives rise to three linked concerns.First,AI can be harnessed to mount sophisticated and far-reaching attacks.Second,as it b

30、ecomes more widely used by financial institutions,more data will be held and transmitted,increasing the attack surface.And third,customer-facing(or application programming interface-accessible)AI systems may be tricked into disclosing information they should not.While cybersecurity defences are adap

31、ting,they may not keep pace with increasingly complex possibly state-sponsored attacks.This suggests a race between benign and malign actors and the cost of protective measures may give larger,systemically important firms an advantage.Market manipulationAI has developed powerful abilities to generat

32、e plausible false or misleading information.This ability may be used to manipulate markets by spreading disinformation,possibly about world events,a companys strength or anything else affecting an assets valuation.AI systems may also be vulnerable to adversarial attacks that involve interfering with

33、 a model(for instance,through the data it interacts with)to misidentify patterns it is later presented with.Some scholars have suggested this could allow bad actors to influence the trades of others in the market,if these are guided by AI models7.Privacy concerns with big dataModern machine learning

34、 models rely on vast amounts of data.Increasingly,much of this data may include sensitive pieces of personal information about people,communities and organisations.AIs predictive capabilities can explain exactly why it is valued by so many financial institutions,who may deploy AI models on their own

35、 proprietary data(such as customer or transaction information).Models might also be trained by developers knowingly or unknowingly on public data that include sensitive personal information,raising legal and ethical problems related to consent,biases,discrimination and individual freedoms.Consider,f

36、or instance,the use of AI to write credit proposals or make credit decisions.In the public sector,Banque de France has experimented with using AI-enabled text-to-data tools to offer better measurements of consumer confidence8.However,sensitive personal data found in the model training dataset may af

37、fect a lending decision or pricing in a way that even the institution is unaware of.Lack of explainabilityA fundamental challenge in managing AI risk today is explainability.Many models particularly those at the frontier of AI development engage in a thinking process that is difficult to understand,

38、even for their creators.This poses challenges in a highly regulated sector such as finance.This issue has been widely cited by regulators as a core challenge to their ability to effectively manage the use of AI in the sector and to develop and enforce guardrails9.The financial sectors increased comp

39、lexity and opacity has been found to play a role in past financial crises by reducing effective oversight and fuelling fear and confusion amid failures10.Research also warns that the integration of black box AI decision-making into parts of the financial system(e.g.in algorithmic trading)may exacerb

40、ate this11.Biases embedded in AI algorithmsLack of explainability also makes model biases harder to discern or correct.Literature on AI biases where datasets or algorithms are skewed unfairly towards particular outcomes,demographics or individuals is bountiful12.This may pose challenges,for example,

41、in credit,real estate and insurance markets.Individuals deemed high risk by AI could face prohibitive pricing or outright denial of service.Regulations may prohibit certain AI increases fraud risk,including by lowering the cost of running phishing-style attacks,which are increasingly hyper-personali

42、sed and effective scams.omfif.org6factors being considered in financial decisions.Yet,unbeknown to the institution,opaque AI models could develop proxy measures that replicate the prohibited factors but remain unethical or illegal.Algorithmic convergence and market instabilityThe potential reliance

43、on AI in trading raises the spectre of portfolio and behavioural convergence.As more investors and firms utilise similar AI models and data,their trades may become increasingly correlated,leading to herding behaviour that could amplify volatility and the risk of flash crashes.The 2010 flash crash,wh

44、ere the Dow Jones Industrial Average plummeted nearly 1,000 points within minutes,serves as a stark reminder of the potential for rapid market disruptions enabled by algorithmic trading.The Securities and Exchange Commission warned that AI deployment in portfolio management may result in more violen

45、t market swings as financial algorithms move together13.Concentration and third-party relianceThe AI supply chain appears increasingly concentrated,including in high-performance chips,cloud computing and AI model development14.This implies a certain degree of concentrated operational risks as well.A

46、I technical talent is also constrained and financial institutions have typically sought to work with large technology partners,who can integrate AI into existing enterprise software and services.For policy-makers and regulators,this presents questions around resiliency and market dynamism.Regulatory

47、 challenges in a rapidly evolving landscapeThere are many challenges to regulating the use of AI in the financial system.These take a number of forms.First,the rapid pace of technological development may outpace regulators ability to respond.Second,the increasing nature of black box AI models and la

48、bs makes it hard for regulators to learn how advanced AI systems are being developed,aligned and deployed.Third,different jurisdictions have different priorities,capacities and exposure to AI,leading to divergences in regulatory attitude and further complications in the oversight of a global system

49、such as finance.Fourth,there is the challenge of regulatory arbitrage,where profitable and risky activity is driven to the least regulated parts of the system,while creating risks for wider market participants.Fifth,there is the compounded challenge of ensuring that AI respects norms or rules in its

50、 behaviour.And finally,governments are facing growing pressure from a perceived AI race,which may have major implications for countries economic competitiveness and national security15.As a consequence,many are likely to be reticent to support regulatory policy responses that they view as imposing o

51、nerous limitations on innovation.There are additional challenges of regulating models that lack explainability in their decision-making.Regardless,achieving AI normative alignment poses a significant challenge for regulators as evaluations to curb bad model behaviour remain methodologically challeng

52、ing.The Securities and Exchange Commission warned that AI deployment in portfolio management may result in more violent market swings as financial algorithms move together.7omfif.orgChapter 2SECOND-ORDER RISKSAnother layer of artificial intelligences risks to finance suggests that the technology may

53、 be widening various forms of inequality.THE ADVENT of artificial intelligence brings with it second-order risks,such as the empirical and theoretical economic effects on the functioning of financial markets or the macroeconomy.Many of these elements may have implications for how the financial secto

54、r fulfils its responsibilities to the public.Particularly germane are the impacts of AI on labour markets,which may spur disequalising effects across and within countries.This may catalyse additional social and political effects,including polarisation.Disrupting insurance economicsThe functioning of

55、 insurance markets can be disrupted by changes to the abilities of buyers and sellers to make predictions.New AI analytical techniques may cause these changes by enabling breakthroughs in personalised medical diagnostics,for instance,which would allow for much more precise identification of individu

56、al risk.If higher-quality predictive information is known to individuals,it could lead to adverse selection where only high-risk individuals seek insurance,making insurance provision un-economic.If the information is known to insurers,more people may be identified as high-risk due to any number of m

57、edical covariates and find themselves less insurable.These effects may necessitate policy intervention,especially in countries where key elements of social provision are insurance-based.Labour market disruption and inequalityPrevious technological change had well-understood implications for labour m

58、arkets and distributional outcomes.Research shows that such change tends to spur labour market polarisation and inequalities16 through at least two mechanisms17.First,labour-saving innovations increase worker productivity,allowing economic TypeRiskDescriptionChallenges for specific financial markets

59、Disrupting insurance economicsAIs ability to improve the quality of predictions may disrupt insurance markets,potentially making insurance more expensive for high-risk individuals or businesses.Wider social and political effectsLabour market disruption and inequalityAI may cause labour market polari

60、sation by displacing middle-wage workers through skill-biased technological change.Erosion of economic indicatorsThe hollowing out of middle-wage work contributes to the erosion of the Phillips curve and a decline in real equilibrium interest rates.Figure 2.Second-order risks of AI in financeSource:

61、OMFIFomfif.org8output to rise faster than wages an Engels pause18.This happened during the Industrial Revolution and seemed to have taken place again in the late 20th centurys digitalisation period.The consequence is higher returns on capital and a decline in labours share of income19.A second mecha

62、nism is skill-biased technological change growth in high-wage labour that is complemented by technological change as well as growth in low-wage labour,but a hollowing out of middle-wage work,which is often displaced by technology.Both theoretical and preliminary empirical evidence suggest that AI ma

63、y be widening inequalities,just as prior technologies did.If AI is productivity-boosting as even conservative estimates suggest it will be it seems likely that there will be an increase in inequalities.That said,these remain early days for AI adoption,and there is no settled consensus yet as to the

64、full labour market effects.Some scholars have suggested that society may need to shift away from one where traditional work is the key mechanism through which people contribute to society20.Erosion of traditional economic indicatorsThese macroeconomic dynamics have considerable implications for the

65、financial sector and the way it serves societal demand.This includes the erosion of the Phillips curve the historically inverse relationship between inflation and unemployment and the decline of real equilibrium interest rates21.Research increasingly suggests that both dynamics owe themselves in par

66、t to the hollowing out of middle-wage work across many western democracies.This has led to an investment-driven economy,whereby access to cheap credit is integral to continued economic expansion.A risk of these dynamics is a growing reliance on speculation in financial markets,due to lower real inte

67、rest rates.This can have self-fulfilling effects on widening wealth inequalities.Both theoretical and preliminary empirical evidence suggest that AI may be widening inequalities,just as prior technologies did.9omfif.orgPOLICY RESPONSES AND FUTURE RESEARCHArtificial intelligence and its increasing in

68、tegration into the financial sector has prompted regulatory bodies worldwide to grapple with the challenges of ensuring responsible and ethical AI adoption.WHILE THE field of AI policy is nascent and regulations specifically tailored to AI in finance remain under development,several countries and in

69、ternational bodies have initiated efforts to address the specific risks and opportunities presented by this technological transformation.The policy landscape surrounding AI in finance is characterised by a balance between fostering innovation and mitigating risks.Governments and regulatory bodies fa

70、ce a complex dilemma:they aim to encourage the development and deployment of AI technologies to maintain competitiveness in the global financial landscape while simultaneously safeguarding against potential systemic risks and ensuring consumer protection.Regional approachesAccording to survey respon

71、ses from Organisation of Economic Co-operation and Development regulators on the state of AI regulation in finance,most regulators believe existing,technology-neutral,principles-based frameworks or pre-existing rules for advanced decision-making models adequately address current AI challenges.Howeve

72、r,many jurisdictions are also developing or considering broader AI-specific legislation.In the US,the regulatory approach towards AI in finance is anchored in existing legal frameworks and principles-based guidance22.This is in line with a broader national AI policy that is increasingly hesitant to

73、enact regulations that might hinder AI development.Meanwhile,the UK has embraced a pro-innovation stance to AI regulation,including in finance23.Although specific AI rules are yet to be established,the Financial Conduct Authority and the Prudential Regulation Authority have articulated their strateg

74、ic approaches to overseeing AI in the financial sector.Canada and Japan have taken similar approaches.The European Union has perhaps gone the furthest in AI-specific regulation with the Artificial Intelligence Act,which aims to create a comprehensive and harmonised framework for AI across various se

75、ctors,including finance24.It adopts a risk-based approach,categorising AI systems based on their potential impact and imposing stricter requirements on high-risk applications.In the context of finance,AI systems used for personal credit scoring,health and life insurance pricing and other critical de

76、cision-making processes have been classified as high-risk.These systems will be subject to stringent requirements regarding transparency,explainability,data quality,human oversight and robustness.An open question in the regulatory landscape is whether AI introduces entirely new categories of risks o

77、r primarily amplifies existing ones.Traditional regulatory frameworks and institutions may not be adequately equipped to address the novel challenges posed by advanced AI technologies,given their complexity,opacity and the swift pace of their development.Areas for further researchAI adoption in fina

78、nce and in the wider economy is expanding rapidly,offering new growth opportunities and efficiencies,while presenting several risks.Addressing the risks posed by AI in and through the finance system will require a flexible and proactive approach between researchers,finance industry professionals,AI

79、developers,regulators and policy-makers.Paths forward for future researchers and policy-makers could include developing a Chapter 3omfif.org10common reporting framework to track AI developments in finance,and AI research focused on model capabilities of particular relevance to risks identified in fi

80、nance.Novel uses of AI in high-risk areas could be deployed in sandboxed environments,to allow innovation while maximising learning,and then be subject to post-deployment monitoring for potential adverse effects.Future research and policy efforts should focus on addressing the risks posed by AI adop

81、tion in the financial sector,with particular attention to both first-order and second-order risks.For first-order risks,a critical area of inquiry is improving AI model interpretability to enhance transparency and accountability in decision-making.Additionally,research is needed to identify the most

82、 effective countermeasures against AI-driven fraud,scams and cybersecurity threats.Given the potential for algorithmic convergence to increase systemic instability,further investigation is required to understand and mitigate these risks.Second-order risks demand an exploration of how AI adoption may

83、 disrupt traditional models for insurance,or other fundamental financial markets such as credit,and what policy interventions could address these challenges.The long-term effects of AI-induced labour market polarisation on financial system stability and socio-economic inequality also warrant closer

84、examination.Furthermore,as AI reshapes the role of labour in the economy,research should explore how the financial sector can adapt to support an economy where traditional employment plays a diminished role.Regulatory and policy challenges present additional areas for study.Future work should focus

85、on designing or fine-tuning regulatory frameworks to balance innovation with robust risk management.International collaborations should also be fostered to address cross-border AI risks in finance.Equally important is understanding how policy-makers can ensure equitable access to AIs benefits in fin

86、ance while not exacerbating existing inequalities.Governments could also look to develop risk mitigation principles and measures at multiple levels,including at the AI model development stage,within financial institutions and through appropriate regulation and policy action.The ethical use of AI in

87、finance also requires greater attention.Research must address how to implement safeguards against privacy violations and unethical data usage in AI model training and deployment.Finally,a focus on future scenarios and strategic planning is key to anticipate and address potential feedback loops or un

88、intended consequences of AI adoption.Such scenario planning should cover a wide spectrum of potential impacts,including operational risks and those implied by concentrated control over AI development by a small number of parties.Public-private collaboration may play a helpful role in fostering resil

89、ience and innovation within AI-enabled financial systems.These areas of inquiry will be critical for ensuring that the financial sector can navigate the challenges of AI while maximising its potential to contribute to a stable,equitable and sustainable economy.An open question in the regulatory land

90、scape is whether AI introduces entirely new categories of risks or primarily amplifies existing ones.11omfif.orgReferences1 Mokyr,J.(2018).The past and the future of innovation:Some lessons from economic history.Explorations in Economic History,69,1326.2 Eisfeldt,A.L.,&Schubert,G.(2024).AI and finan

91、ce(Working Paper No.33076).National Bureau of Economic Research.3 Kelly,B.,&Xiu,D.(2023).Financial machine learning.Foundations and Trends in Finance,13(34),205236.4 Hall,J.(2024,May 7).Monsters in the deep?Speech presented at the University of Exeter Business School.5 University of Waterloo.(2023).

92、How secure are voice authentication systems really?Attackers can break voice authentication with up to 99 percent success within six tries.6 Walsh,M.(2024,October 28).A framework for detection in an era of rising deepfakes.7 Goldstein,J.,et al.(2023,March 7).Generative AI is enabling fraud and misin

93、formation-Here is what you should know.Center for Security and Emerging Technology.8 Bank for International Settlements.(2024).Artificial intelligence and the economy:Implications for central banks(BIS Annual Economic Report,25 June 2024).Bank for International Settlements.9 Crisanto,J.C.,Leuterio,C

94、.B.,Prenio,J.,&Yong,J.(2024).Regulating AI in the financial sector:Recent developments and main challenges(FSI Insights on Policy Implementation No.63).Financial Stability Institute.10 Turner,A.(2015).Between debt and the devil:Money,credit,and fixing global finance.Princeton University Press.11 Bat

95、haee,Y.(2018)The artificial intelligence black box and the failure of intent and causation.Harvard Journal of Law&Tech,31(2)12 Jarrell,N.S.,McGrath,S.,Ferguson Edwards,S.,&Nagarajan,J.(2023,March 17).How to mitigate AI discrimination and bias in financial services.Ernst&Young LLP.13 Vereckey,B.(2022

96、,October 12).SECs Gary Gensler on how artificial intelligence is changing finance.MIT Sloan Management Review14 Vipra,J.,&Korinek,A.(2023).Market concentration implications of foundation models:The invisible hand of ChatGPT.Brookings Institution.15 Goujon,R.(2024,December 27).The real stakes of the

97、AI race:What America,China,and middle powers stand to gain and lose.Foreign Affairs 16 Autor,D.H.,&Dorn,D.(2013).The growth of low-skill service jobs and the polarization of the US labor market.American Economic Review,103(5),15531597.17 Allen,R.C.(2009).Engels pause:Technical change,capital accumul

98、ation,and inequality in the British industrial revolution.Explorations in Economic History,46(4),418435.18 Acemoglu,D.,&Restrepo,P.(2022).Tasks,automation,and the rise in US wage inequality.Econometrica,90(5),19732016.19 Susskind,D.(2024)We must change the nature of growth.Finance&Development Magazi

99、ne,International Monetary Fund.20 Friedrich,C.,&Selcuk,P.(2022).The impact of globalization and digitalization on the Phillips curve(Bank of Canada Staff Working Paper,No.2022-7).Bank of Canada.21 OECD.(2024).Regulatory approaches to artificial intelligence in finance(OECD Artificial Intelligence Pa

100、pers,No.24).OECD Publishing.22 US Department of the Treasury.(2024).Managing artificial intelligence-specific cybersecurity risks in the financial services sector.23 Bank of England&Financial Conduct Authority.(2024,November 21).Artificial intelligence in UK financial services-2024.24 Ilg,B.C.(2024)

101、.AI in finance:How does the uptake of Artificial Intelligence systems impact finance?Directorate-General for Financial Stability,Financial Services and Capital Markets Union Newsletteromfif.org12Further readingAcemoglu,D.(2002).Technical change,inequality,and the labor market.Journal of Economic Lit

102、erature,40(1),772.Acemoglu,D.(2024).The simple macroeconomics of AI.Economic Policy,39(120),31173150.Acemoglu,D.,&Restrepo,P.(2018).Artificial intelligence,automation,and work(Working Paper No.24196).National Bureau of Economic Research.Acemoglu,D.,&Restrepo,P.(2020).Unpacking skill bias:Automation

103、and new tasks(No.w26681).National Bureau of Economic Research.Aderemi,S.,Olutimehin,D.O.,Nnaomah,U.I.,Orieno,O.H.,Edunjobi,T.E.,&Babatunde,S.O.(2024).Big data analytics in the financial services industry:Trends,challenges,and future prospects:A review.International Journal of Science and Technology

104、Research Archive,6(1),147166.Aldasoro,I.,Gambacorta,L.,Korinek,A.,Shreeti,V.,&Stein,M.(2024).Intelligent financial system:How AI is transforming finance(BIS Working Paper No.1194).Bank for International Settlements.Allen,R.C.(2009).Engels pause:Technical change,capital accumulation,and inequality in

105、 the British industrial revolution.Explorations in Economic History,46(4),418435.Ari,A.,Garcia-Macia,D.,&Mishra,S.(2023,May 12).Has the Phillips curve become steeper?International Monetary Fund.Autor,D.(2022).The labor market impacts of technological change:From unbridled enthusiasm to qualified opt

106、imism to vast uncertainty(Working Paper No.30074).National Bureau of Economic Research.Autor,D.(2024).Applying AI to rebuild middle class jobs(NBER Working Paper No.32140).National Bureau of Economic Research.Bahoo,S.,Cucculelli,M.,Goga,X.,&Mondolo,J.(2024).Artificial intelligence in finance:A compr

107、ehensive review through bibliometric and content analysis.SN Business&Economics,4(23).Baily,M.N.,&Elliott,D.J.(2013).The role of finance in the economy:Implications for structural reform of the financial sector.Bai,Y.et al.(2022)Constitutional AI:Harmlessness from AI Feedback Bateman,J.(2020).Deepfa

108、kes and synthetic media in the financial system:Assessing threat scenarios.Belenguer L.(2022).AI bias:exploring discriminatory algorithmic decision-making models and the application of possible machine-centric solutions adapted from the pharmaceutical industry.AI Ethics.Bengio,Y.,Hinton,G.,Yao,A.,So

109、ng,D.,Abbeel,P.,Darrell,T.,Harari,Y.N.,Zhang,Y.Q.,Xue,L.,.,&Mindermann,S.(2024).Managing extreme AI risks amid rapid progress:Preparation requires technical research and development,as well as adaptive,proactive governance.Science,384(6698),842845.Bernanke,B.(19921993).Credit in the macroeconomy.Fed

110、eral Reserve Bank of New York.Boehmer,E.,Fong,K.,&Wu,J.(2020).Algorithmic trading and market quality:International evidence.Journal of Financial and Quantitative Analysis.Published online by Cambridge University Press on October 13,2020.Boehmer,E.,Fong,K.,&Wu,J.(n.d.).Algorithmic trading and market

111、quality:International evidence.Boukherouaa,E.B.,AlAjmi,K.,Deodoro,J.,Farias,A.,&Ravikumar,R.(2021).Powering the digital economy:Opportunities and risks of artificial intelligence in finance.Brainard,L.(2018,November 13).What are we learning about artificial intelligence in financial services?Speech

112、presented at Fintech and the New Financial Landscape,Philadelphia,Pennsylvania.erneviien,J.,&Kabainskas,A.(2024).Explainable artificial intelligence(XAI)in finance:A systematic literature review.Artificial Intelligence Review,57,Article 216.Colliard,J.-E.,Foucault,T.,&Lovo,S.(2023).Algorithmic prici

113、ng and liquidity in securities markets.Columbia University,School of International and Public Affairs.(2023).Deepfakes and disinformation in the finance sector:Strategies to prevent and deter.Das,R.,&Sandhane,R.(2021).Artificial intelligence in cyber security.Journal of Physics:Conference Series,196

114、4,042072.Fan,J.,Liu,Q.,Song,Y.,&Wang,Z.(2024).Measuring misinformation in financial markets.Financial Stability Board.(2024).The Financial Stability Implications of Artificial Intelligence.Fong,B.(2021).Analysing the behavioural finance impact of fake news phenomena on financial markets:A representa

115、tive agent model and empirical validation.Financial Innovation,7,Article 53.Gallagher,M.,Pitropakis,N.,Chrysoulas,C.,Papadopoulos,P.,Mylonas,A.,&Katsikas,S.(2022).Investigating machine learning attacks on financial time series models.Computers&Security,123,Article 102933.Godoy de Araujo,D.K.,Doerr,S

116、.,Gambacorta,L.,&Tissot,B.(2024).Artificial intelligence in central banking(BIS Bulletin No.84).Bank for International Settlements.Grout,P.A.(2021).AI,ML,and competition dynamics in financial markets.Oxford Review of Economic Policy,37(3),618635.Gruetzemacher,R.,&Whittlestone,J.(2022).The transforma

117、tive potential of artificial intelligence.Futures,135,102884.13omfif.orgInternational Monetary Fund.Monetary and Capital Markets Department.(2024).Advances in artificial intelligence:Implications for capital market activities.In International Monetary Fund Report(Chapter 3).International Organizatio

118、n of Securities Commissions.(2021).The use of artificial intelligence and machine learning by market intermediaries and asset managers.Jada,I.,&Mayayise,T.O.(2024).The impact of artificial intelligence on organisational cyber security:An outcome of a systematic literature review.Data and Information

119、 Management,8(2),Article 100063.Johnson,K.,Pasquale,F.,&Chapman,J.(2019).Artificial intelligence,machine learning,and bias in finance:Toward responsible innovation.Fordham Law Review,88(2),Article 5.Kaur,R.,Gabrijeli,D.,&Klobuar,T.(2023).Artificial intelligence for cybersecurity:Literature review an

120、d future research directions.Information Fusion,97,Article 101804.Klein,A.(2020,July 10).Reducing bias in AI-based financial services.Brookings Institution.Klein,T.,&Walther,T.(2024).Advances in explainable artificial intelligence(xAI)in finance.Finance Research Letters,70,Article 106358.Kogan,I.(20

121、18)Could your next analyst be a computer?Psychoanalysis in the digital era.Psychoanalytic Perspectives on Virtual Intimacy and Communication in Film.1st Edition.Routledge.Korinek,A.(2024).Economic policy challenges for the age of AI(NBER Working Paper No.32980).National Bureau of Economic Research.L

122、ancieri,F.,Edelson,L.,&Bechtold,S.(2024).AI regulation:Competition,arbitrage®ulatory capture.SSRNLevine,R.(2021).Finance,growth,and inequality(IMF Working Paper).International Monetary Fund.Malkiel,B.G.(2003).The efficient market hypothesis and its critics(CEPS Working Paper No.91).Princeton Univ

123、ersity,Center for Economic Policy Studies.Martin,K.D.,&Zimmermann,J.(2024).Artificial intelligence and its implications for data privacy.Current Opinion in Psychology,58,Article 101829.Mittermaier,M.,Raza,M.M.,&Kvedar,J.C.(2023).Bias in AI-based models for medical applications:Challenges and mitigat

124、ion strategies.npj Digital Medicine,6,Article 113.Murikah,W.,Nthenge,J.K.,&Musyoka,F.M.(2024).Bias and ethics of AI systems applied in auditing:A systematic review.Scientific African,25,Article e02281.Muro,M.,&Liu,S.(2021,September).The geography of AI:Which cities will drive the artificial intellig

125、ence revolution?Brookings Institution.Muro,M.,Jacobs,J.,&Liu,S.(2023,July 20).Building AI cities:How to spread the benefits of an emerging technology across more of America.Brookings Institution.Mustak,M.,Salminen,J.,Mntymki,M.,Rahman,A.,&Dwivedi,Y.K.(2023).Deepfakes:Deceptions,mitigations,and oppor

126、tunities.Journal of Business Research,154,Article 113368.Olaiya,O.P.,Cynthia,A.C.,Usoro,S.O.,Obani,O.Q.,Nwafor,K.C.,&Ajayi,O.O.(2024).The impact of big data analytics on financial risk management.International Journal of Science and Research Archive,12(2),821827.Rice,L.,&Swesnik,D.(2012,June 67).Dis

127、criminatory effects of credit scoring on communities of color.Paper presented at the Symposium on Credit Scoring and Credit Reporting,Suffolk University Law School and National Consumer Law Center.Rothschild,M.,&Stiglitz,J.(1976).Equilibrium in competitive insurance markets:An essay on the economics

128、 of imperfect information.The Quarterly Journal 1 of Economics,90(4),629649.2 Salem,A.H.,Azzam,S.M.,Emam,O.E.,&Abohany,A.A.(2024).Advancing cybersecurity:A comprehensive review of AI-driven detection techniques.Journal of Big Data,11,Article 105.Sandoval,M.-P.,de Almeida Vau,M.,Solaas,J.,&Rodrigues,

129、L.(2024).Threat of deepfakes to the criminal justice system:A systematic review.Crime Science,13,Article 41.Shivampeta,P.(2023).Artificial intelligence for cyber security threats(Masters thesis).Governors State University.Smith,L.T.,Pirchalski,E.,&Golbin,I.(2022).Avoiding unfair bias in insurance ap

130、plications of AI models.Society of Actuaries.Stetler,C.(2024,April 10).AI usage in the insurance industry:Future or fad.Turner,A.(2015).Between debt and the devil:Money,credit,and fixing global finance.Princeton University Press.U.S.Department of the Treasury.(2024).Artificial intelligence:Report on

131、 the uses,opportunities,and risks of artificial intelligence in the financial services sector.Weber,P.,Carl,K.V.,&Hinz,O.(2024).Applications of explainable artificial intelligence in financea systematic review of finance,information systems,and computer science literature.Management Review Quarterly,74,867907.Yuferova,D.(2024).Algorithmic trading and market efficiency around the introduction of the NYSE Hybrid Market.Journal of Financial Markets,69.Digital Monetary Institute