OCDE - Supervision of AI in finance Challenges

SUPERVISION

OF ARTIFICIAL

INTELLIGENCE IN

FINANCE

CHALLENGES, POLICIES AND

PRACTICES

OECD ARTIFICIAL INTELLIGENCE PAPERS January 2026 No. 542 

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

This work is published under the responsibility of the Secretary -General of the OECD. The opinions expressed and arguments employed herein do not necessarily reflect the official views of the Member countries of the OECD. This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area.

Photo credits: © Kjpargeter/Shutterstock

© OECD 2026

Attribution 4.0 International (CC BY 4.0) This work is made available under the Creative Commons Attribution 4.0 International licence. By using this work, you accept to be bound by the terms of this licence (https://creativecommons.org/licenses/by/4.0/). Attribution – you must cite the work. Translations – you must cite the original work, identify changes to the original and add the following text: In the event of any discrepancy between the original work and the translation, only the text of original work should be considered valid. Adaptations – you must cite the original work and add the following text: This is an adaptation of an original work by the OECD. The opinions expressed and arguments employed in this adaptation should not be reported as representing the official views of the OECD or of its Member countries. Third-party material – the licence does not apply to third-party material in the work. If using such material, you are responsible for obtaining permission from the third party and for any claims of infringement. You must not use the OECD logo, visual identity or cover image without express permission or suggest the OECD endorses your use of the work.

Third-party material – the licence does not apply to third-party material in the work. If using such material, you are responsible for obtaining permission from the third party and for any claims of infringement. You must not use the OECD logo, visual identity or cover image without express permission or suggest the OECD endorses your use of the work. Any dispute arising under this licence shall be settled by arbitration in accordance with the Permanent Court of Arbitration (PCA) Arbitration Rules 2012. The seat of arbitration shall be Paris (France). The number of arbitrators shall be one.

Disclaimers 3

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

While most OECD Member countries consider they have appropriate regulations for the use of AI in finance, challenges may arise in the interpretation and implementation of applicable AI regulations by financial supervisors. This paper analyses current supervisory approaches to the use of AI in finance and challenges in overseeing its adoption. The paper also reviews supervisory practices that balance promoting responsible AI adoption in finance with policy objectives of financial market stability and integrity, and the protection of financial consumers. This paper is part of the series “OECD Artificial Intelligence Papers”, https://doi.org/10.1787/dee339a8-en

Abstract4 

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

Table of contents Abstract 3 Acknowledgements 5 Executive summary 6 1 Translating policies into effective oversight for AI in finance 9 1.1. Supervisory approaches to AI in finance 9 1.2. Oversight frameworks: interplay between sectorial rules and other policies 11 1.3. Data gaps and monitoring tools 13 2 Potential challenges to supervision of AI in finance 16 2.1. Reported challenges to AI supervision in finance 16 2.2. Model risk management, validation and compliance assessment 17 2.3. Explainability, transparency and fairness 20 2.4. Governance and data management 22

2 Potential challenges to supervision of AI in finance 16 2.1. Reported challenges to AI supervision in finance 16 2.2. Model risk management, validation and compliance assessment 17 2.3. Explainability, transparency and fairness 20 2.4. Governance and data management 22 3 Supervisory practices to balance innovation and stability 24 3.1. Consider carefully calibrated additional guidance on supervisory expectations/ supervisory guidance when this is warranted 25 3.2. Encourage public-private cooperation, leveraging sandboxes and novel AI model testing 26 3.3. Invest in supervisory capacity, upskilling and use of AI-driven SupTech 29 3.4. Encourage policymaker coordination across sectors and jurisdictions 30 3.5. Evolution of AI Supervision in finance: pushing the boundaries of tech neutrality? 31 References 32 Notes 36

FIGURES Figure 1.1. Examples of areas covered by existing financial sector rules 10 Figure 1.2. Supervisory coordination efforts 13 Figure 2.1. Identified challenges in supervision of AI in finance 16 Figure 3.1. Appropriate AI regulation is reported to be in place to address areas related to supervisory challenges 24 Figure 3.2. Clarifications around the applicability of existing rules/ regulations/ other policy frameworks on AI applications in finance 25 Figure 3.3. Action to promote the upskilling of supervisors in relation to AI ongoing in majority of OECD countries 29 5

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

Acknowledgements This report provides analysis of current supervisory approaches to Artificial Intelligence (AI) in Finance and discusses reported challenges encountered in the interpretation and implementation of applicable AI regulations by financial supervisors in some jurisdictions. The report discusses polic ies and supervisory practices that balance the promotion of responsible adoption of AI in finance with policy objectives of stability and integrity of financial markets and protection of financial consumers.

discusses reported challenges encountered in the interpretation and implementation of applicable AI regulations by financial supervisors in some jurisdictions. The report discusses polic ies and supervisory practices that balance the promotion of responsible adoption of AI in finance with policy objectives of stability and integrity of financial markets and protection of financial consumers. The report has been developed by the Capital Markets and Financial Institutions of the OECD Directorate for Financial and Enterprise Affairs. It was drafted by Iota Kaousar Nassr under the supervision of Fatos Koc, Head of the Financial Markets Unit, and Serdar Çelik, Head of Division. Eva Abbott, Liv Gudmundson and Mathilde Le Pichon provided editorial and communication support. The authors gratefully acknowledge valuable input and feedback provided by the following individuals and organisations: Sara G. Castellanos, Banco de México; Giuseppe Grande, Banca d’Italia; Mikari Kashima, Bank of Japan; Rohan Paris and Paull Randt, U.S. Department of the Treasury; Jasmine Tan, Australian Securities and Investments Commission. The report was discussed by the OECD Committee on Financial Markets , chaired by Mr Seiichi Shimizu, Assistant Governor, Bank of Japan, o n 11 September 2025. The report constitutes part of the horizontal OECD project on Artificial Intelligence.6 

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

Executive summary The transformative potential of artificial intelligence (AI) innovation, catalysed by advancements in generative AI (GenAI) and large language models (LLMs), is poised to significantly reshape the global financial sector. The finance sector , having leveraged machine learning [ML] models for decades , is progressively exploring and deploying GenAI models, while also exploring Agentic AI capabilities. In OECD economies existing regulatory requirements remain applicable irrespective of the technology used to deliver a financial service or product, given the technology -neutral principle guiding financial regulation. While regulation provides the foundational legal architecture for financial oversight, supervision is a dynamic and ongoing process through which rules and policies are interpreted in practice to ensure

In OECD economies existing regulatory requirements remain applicable irrespective of the technology used to deliver a financial service or product, given the technology -neutral principle guiding financial regulation. While regulation provides the foundational legal architecture for financial oversight, supervision is a dynamic and ongoing process through which rules and policies are interpreted in practice to ensure compliance and to identify and assess emerging risks to the integrity and stability of the markets. Financial supervision therefore serves as the practical enforcement mechanism of financial regulation, ensuring that policies translate into effective oversight and resilient financial markets . It is at this level of practical interpretation and implementation of AI policies in finance that challenges may arise , given the intrinsic characteristics of AI innovation, particularly advanced forms of AI. This paper analyses current supervisory approaches and examines reported challenges encountered in the supervision of AI in finance by some countries . It also discusses possible supervisory practices that balance the promotion of the responsible adoption of AI in finance with policy objectives related to the stability and integrity of financial markets and the protection of financial consumers. The report builds on earlier work of the Committee on Financial Markets on Regulatory Approaches to AI in Finance and draws on input from the 2024 OECD Survey on AI in finance and subsequent contributions. Supervisory approaches to AI in finance Despite differences in supervisory approaches (such as reliance on legacy frameworks in some regions or the development of AI -specific guidelines and supervisory expectations in others) supervisory efforts are anchored in common principles, notably a risk-based and technology-neutral approach. Where new policy frameworks have been introduced explicitly for AI in finance, layering AI-specific frameworks on top of preexisting sector-specific rules may complicate the application of supervisory mandate. Efforts are therefore needed to promote and pursue streamlining and simplification of regulation , as well as clarity and consistency in supervisory interpretation , where necessary . This includes identifying any overlaps, conflicts or inconsistencies in regulations applicable to AI , and clarifying the interpretation of these rules for the purposes of AI supervision, if and where necessary, with a view to assisting supervised entities in their compliance efforts. 7

overlaps, conflicts or inconsistencies in regulations applicable to AI , and clarifying the interpretation of these rules for the purposes of AI supervision, if and where necessary, with a view to assisting supervised entities in their compliance efforts. 7

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

Reported challenges in the supervision of AI in finance Some of the most prominent reported challenges in the supervision of AI in finance relate to distinctive characteristics of AI innovation, such as the pace of its evolution (at least thus far), the opaqueness and complexity of the underlying technology, and novelties related to its dynamic n ature and the potential ly high degree of autonomy it could , in theory, entail. The lack of comprehensive data on AI adoption by financial services firms complicates the assessment of its use and may pose challenges for monitoring associated vulnerabilities. This is further exacerbated by the growing significance of non -supervised entities, such as third-party technical vendors, many of which operate outside the scope of formal oversight by financial regulators. Challenges reported by authorities largely mirror the compliance challenges identified by supervised entities, which may in turn impede the wider deployment of AI by financial firms. These challenges include model risk management, validation and compliance assessment of increasingly complex AI systems , as well as limitations related to model explainability. They also include limited transparency and associated considerations around assessing robustness and upholding the fairness of model outputs; alongside governance and data -management challenges . For example, articulating how the concept of ‘human in the loop’ should apply in practice, depending on the context, is reported as challenging by some authorities. While existing requirements continue to apply and supervised entities are expected to take AI -specific aspects into account and adapt their risk -management frameworks accordingly, potential guidance and clarification on how compliance requirements align with advanced AI models’ technical specificities could be beneficial in some jurisdictions . Depending on the case, such guidance could address any perceived ambiguity as to the way model risk management frameworks should be interpreted and operationalised

aspects into account and adapt their risk -management frameworks accordingly, potential guidance and clarification on how compliance requirements align with advanced AI models’ technical specificities could be beneficial in some jurisdictions . Depending on the case, such guidance could address any perceived ambiguity as to the way model risk management frameworks should be interpreted and operationalised given, for example, the lack of explainability and the dynamic adaptability and recalibrat ion of AI models. Rather than imposing rigid or overly prescriptive requirements that could inadvertently hinder the adoption of AI innovation, it may be more effective to consider providing interpretative guidance and practical clarifications on the application of existing model risk management frameworks in AI contexts if and when this is deemed necessary. Balancing policy objectives with the promotion of the responsible adoption of AI In jurisdictions where supervised entities report challenges arising from a perceived lack of clarity, particularly in light of the overlay of newly adopted AI regulation, carefully calibrated guidance on the interpretation of high-level principles could be beneficial . Additional clarifications could help provide legal certainty for firms, which in turn may strengthen confidence and encourage further investment in responsible AI innovation. Greater clarity for the finance industry regarding regulatory requirements and how to meet them through supervisory expectations and guidance could be beneficial in such jurisdictions. Such guidance could help market participants ensure regulatory compliance and reduce perceived regulatory uncertainty, and support more effective oversight and consistent regulatory outcomes. Any guidance provided should be very carefully designed and calibrated, to avoid a negative effect on AI adoption by impeding firms ’ ability to flexibly explore using new technolog ies. Overly prescriptive approaches should be avoided , given the rapid pace of technological innovation , and a risk -based approach may be most effective for enabling financial institutions to address key risks where needed. Enhanced forms of proactive engagement between supervisors and industry stakeholders , beyond standard supervisory activities, could foster mutual understanding. Close and sustained engagement with industry can yield significant benefits for supervised entities, while also improving authorities’ understanding of the challenges encountered by supervised entities in their compliance efforts. Proactive engagement with the industry through AI-specific testing , such as sandboxes1 or model testing, can

standard supervisory activities, could foster mutual understanding. Close and sustained engagement with industry can yield significant benefits for supervised entities, while also improving authorities’ understanding of the challenges encountered by supervised entities in their compliance efforts. Proactive engagement with the industry through AI-specific testing , such as sandboxes1 or model testing, can provide the confidence and clarity needed to encourage innovation while protecting markets and their8 

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

participants and safeguarding stability. Novel initiatives involving model testing can cultivate productive dialogue between firms developing or deploying AI models and supervisory bodies, fostering mutual understanding and supporting model validation (e.g. the UK Financial Conduct Authority AI Live Testing). Increased capacity and upskilling of financial supervisors will also be necessary to achieve monitoring and oversight objectives, as well as to enable authorities to develop and deploy AI as part of the supervisory activity inter alia through SupTech tools that incorporate AI innovation . Co-ordinated efforts among supervisory authorities could enable the strategic pooling of expertise and institutional capacity in relation to AI-based SupTech tools. Investment is also required to conduct further research into the potential longterm impacts of AI on financial market structures, competition, and financial stability. Maintaining a flexible, agile and adaptive stance to the financial supervision at the practical level could help address the supervisory challenges discussed above while allowing oversight to keep pace with technological advances. In some cases, allowing for technology -specific guidance or the consideration of novel methodologies and techniques to enrich the supervisory toolkit could assist supervisors in achieving a balance between fostering innovation and ensuring stability. Public-private dialogue between supervisors and regulated entities should accompany the consideration of new supervisory methods, techniques, and tools. Continuous assessment of the supervisory landscape is necessary to ensure that it remains fit for purpose. Supervision should also remain open to enriching and adapting this framework to reflect the realities and specific characteristics of AI innovation, in order to foster responsible AI innovation while mitigating associated risks. 9

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

purpose. Supervision should also remain open to enriching and adapting this framework to reflect the realities and specific characteristics of AI innovation, in order to foster responsible AI innovation while mitigating associated risks. 9

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

1.1. Supervisory approaches to AI in finance Regulation provides the foundational legal architecture for financial oversight. OECD analysis indicates that most jurisdictions consider they have appropriate regulation in place for the use of AI in finance (OECD, 2024 [1]). This includes pre -existing regulation, newly introduced product regulation or crosssectorial rules, as well as non -binding policy guidance and national policy frameworks which either specifically target financial activities or apply across sectors, incl uding finance. It is also important to note that the vast majority of responding jurisdictions do not plan to introduce new regulations for AI in finance. Given the technology neutral principle guiding financial regulation, existing rules and guidance remain applicable regardless of the underlying technology used to deliver a financial service or product . This includes laws and regulations on prudent business practices, consumer and investor protection, cybersecurity, and operational resilience, among other areas (see Box 2.1). Many of the risks related to AI are not necessarily new or unique to AI innovati on but are instead exacerbated or amplified by the us e of such innovation, or manifest in different ways (OECD, 2021[2]; 2023[3]). Advances in technology do not render existing safety and soundness standards or compliance requirements obsolete. This is particularly true in the financial sector, where the use of models has long been integral to the business strategies of market participants, spanning several decades. Financial supervision operates alongside financial regulation to ensure compliance, stability, and integrity within the financial system and goes beyond the simple enforcement of regulations to include the management of risks . While financial regulation establishes the rules and standards that financial institutions must follow, supervision actively monitors, assesses, and enforces these requirements . Financial supervision therefore serves as the practical enforcement mechanism of financial regulation,

within the financial system and goes beyond the simple enforcement of regulations to include the management of risks . While financial regulation establishes the rules and standards that financial institutions must follow, supervision actively monitors, assesses, and enforces these requirements . Financial supervision therefore serves as the practical enforcement mechanism of financial regulation, ensuring that policies translate into effective oversight and resilient financial markets . It is at this level of practical interpretation of AI policies in finance that supervisory challenges may arise. Supervisory approaches to AI in finance vary across jurisdictions , ranging from leveraging existing principles-based frameworks ( e.g. UK’s framework ) to developing AI -specific guidance (e.g. Monetary Authority of Singapore FEAT framework) and integrating cross -sectoral AI regulat ory requirements into certain areas of financial supervision (e.g. EU AI Act). Indicatively, in the UK, authorities primarily rely on established principles -based frameworks to guide oversight. Singapore’s Monetary Authority has developed dedicated AI governance principles tailored to guide the sector into addressing specific challenges posed by AI through enhanced governance . In the EU, the AI Act incorporates AI-specific requirements for use cases deemed high -risk in insurance and banking, within a broader cross-sectoral regulation, which will need to be incorporated into existing supervisory strategies .

1 Translating policies into effective oversight for AI in finance10 

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

Box 1.1. Regulatory approaches to AI in finance In September 2024, the Committee on Financial Markets released an overview analysing different regulatory approaches to the use of AI in finance in 49 OECD and non -OECD jurisdictions based on a dedicated Survey on Regulatory Approaches to AI in Finance. The OECD analysis provides examples of rules and regulations that may apply to the use of AI in finance and that can be grouped under a set of areas as depicted in Figure 1.1. Most of the covered areas relate to risk management and, in particular, model risk management; data -related frameworks;

The OECD analysis provides examples of rules and regulations that may apply to the use of AI in finance and that can be grouped under a set of areas as depicted in Figure 1.1. Most of the covered areas relate to risk management and, in particular, model risk management; data -related frameworks; consumer and investor protection, as well as governance and accountability requirements. Figure 1.1. Examples of areas covered by existing financial sector rules

Note: Non-exhaustive, as reported by respondents to the survey.

Source: OECD (2024[1]), Regulatory approaches to Artificial Intelligence in finance, https://www.oecd.org/en/publications/regulatoryapproaches-to-artificial-intelligence-in-finance_f1498c02-en.html Across jurisdictions, different forms of binding and/or non -binding policy instruments have emerged to complement existing financial regulations in response to AI advances. Some countries have enacted cross-sectoral legislation encompassing financial activities (e.g. EU AI Act and national laws in Brazil, Colombia, and Peru ), while others have pursued targeted regulatory proposals focused on specific actors and activities . In parallel, about c. a quarter of respondents to the OECD survey have issued non-binding guidance, including blueprints, principles, and white papers, either at the cross -sectoral level or tailored to financial domains. These instruments generally aim to establish priorities and promote safe and responsible AI innovation. Despite variatio ns in format, there are significant commonalities in content, emphasising fairness, accountability, ethical use, compliance, transparency, and robust governance mechanisms. Some jurisdictions additionally urge financial authorities to leverage their full supervisory remit to mitigate AI-related risks to consumers and investors. Importantly, the different approaches outlined above are not mutually exclusive.

Source: OECD (2024[1]), Regulatory approaches to Artificial Intelligence in Finance,

https://www.oecd.org/content/dam/oecd/en/publications/reports/2024/09/regulatory-approaches-to-artificial-intelligence-infinance_43d082c3/f1498c02-en.pdf

0 5 10 15 20 25 Competition Incident reporting/ Liability Explainability/ interpretability Prudential Market integrity/ market conduct Operational resilience ICT management Outsourcing/ third party risk Ethical/ human rights (incl. anti-discrimination) Governance Cyber-risk Disclosure Investor/ Consumer protection Model risk management Data protection/ privacy Risk management Number of Respondents 11

SUPERVISION OF ARTIFICIAL INTELLIGENCE IN FINANCE © OECD 2026

Although supervisory approaches to AI in finance may appear disparate across jurisdictions, they are all underpinned by the same foundational principles, in particular the risk-based approach to supervision and a technology neutral/agnostic stance to innovation. Under risk-based supervision, supervisory resources and interventions are prioritised according to the relative risk profile of financial institutions or sectors, with the intensity of supervision aligned to the prevailing risks to which these are exposed. Rather than applying uniform oversight, risk-based supervision enables supervisors to focus more intensively on entities or activities that pose higher risks to financial stability, consumer protection, or market integrity. In addition to enhancing the effectiveness of supervisory efforts by tailor ing their monitoring, inspection, and enforcement strategies accordingly, risk-based supervision can also support financial inclusion by reducing regulatory burdens on lower-risk entities or activities. Technology neutrality, which serves as a foundational element of financial regulation in OECD countries, is also reflected in oversight practices. Financial supervisors aim to apply consistent oversight standards irrespective of the technologies employed, thereby ensuring that regulation remains resilient and adaptable in the face of innovation. When AI is used in areas covered by existing rules or guidance, such rules or guidance should generally apply, whether decision s are made by AI (with or without hum an intervention),

irrespective of the technologies employed, thereby ensuring that regulation remains resilient and adaptable in the face of innovation. When AI is used in areas covered by existing rules or guidance, such rules or guidance should generally apply, whether decision s are made by AI (with or without hum an intervention), traditional models, or humans (OECD, 2024[1]). This transposition enables regulatory frameworks to foster competition and innovation without prescribing or favoring specific technological solutions, while maintaining their core objectives of market integrity, consumer protection, and financial stabilit y. Nevertheless, at the level of practical implementation, financial supervisors have reported tangible challenges in effectively translating and interpretati ng technology-neutral regulatory provisions across specific domains associated with the use of AI in finance (Section 4.1). These challenges frequently stem from the distinctive characteristics and novel dimensions of AI innovation, particularly its increasing complexity and rapid pace of evolution. Supervisory challenges are also reported as resulting fr om the interplay between sectorial rules and new regulation or specific AI -related guidance, where these have been established. Additional challenges relate to evolving institutional and market structures, for example, due to the growing role of technology providers in the deployment and scaling of AI by financial market participants. Additionally, the limited availability of granular data on the current state of AI adoption can further complicate monitoring activities and may hinder the effective management of identified risks. 1.2. Oversight frameworks: interplay between sectorial rules and other policies Supervisory challenges may arise from the interplay between existing tech-neutral sectoral regulations and emerging cross-sectorial or finance-specific AI policy frameworks, where such rules have been formulated. Depending on the jurisdiction, oversight of financial activities involving the use of AI could require the interpretation of a combination of pre -existing financial sector regulation, c ross-sectorial policies and practices (e.g. anti -discrimination practices and ethics -related rules such as the US Interagency Fair Lending Examination Procedures (FFIEC, 2009 [4]), product -safety or other cross -sectorial regulation

practices (e.g. anti -discrimination practices and ethics -related rules such as the US Interagency Fair Lending Examination Procedures (FFIEC, 2009 [4]), product -safety or other cross -sectorial regulation introduced explicitly for AI with applicable financ ial use cases (e.g. EU AI Act (EU, 2024[5])), and newly introduced sectorial rules, guidance or principles for (parts of) the financial sector (e.g. guidance on data ethics within the insu