Artificial Intelligence for Financial Sector Supervision: An Emerging Market and Developing Economies Perspective
World Bank - 01/2025
- AI adoption in financial sector supervision within Emerging Market and Developing Economies (EMDEs) is nascent, with financial institutions, particularly fintechs, generally ahead of regulatory authorities.
- Most EMDE financial sector authorities view AI adoption as a board-level priority, anticipating a net positive impact on efficiency, competition, and inclusion over the next five years.
- While basic Generative AI (GenAI) tools are being used by staff for tasks like drafting and summarization, core supervisory AI applications are still in early stages, with many authorities conducting tests and pilot programs.
- Key challenges to AI adoption for supervision in EMDEs include data privacy and security, internal skills gaps and workforce readiness, AI model complexities, and integrating AI into existing technological infrastructure.
- African EMDE authorities specifically highlight internal skills gaps and integration challenges as top barriers, with a significant portion lacking expertise to assess vendor-provided AI tools.
- Data quality and availability are critical limitations, with many EMDE authorities facing fragmented or unstructured data that requires substantial preparatory work for AI model training and validation.
- Integrating AI into legacy IT systems is a major hurdle, as these environments are often siloed and lack the interoperable data structures required by modern AI tools.
- Cloud infrastructure offers potential benefits for AI adoption, but legal and regulatory constraints, data sovereignty concerns, and vendor dependency pose significant challenges for EMDE authorities.
- Legal and regulatory hurdles related to data privacy, security, and localization are critical concerns, impacting authorities' ability to leverage AI tools that require extensive data access and processing.
- Cybersecurity risks are paramount, with AI applications increasing the attack surface and the potential for data leakage or theft, particularly in African EMDEs where cybersecurity frameworks may need strengthening.
- AI model challenges, including lack of transparency, explainability (the "black box" problem), and potential for inaccurate outputs or hallucinations, are significant risks, with 70% of respondents citing lack of transparency as a top risk.
- Operational risks, such as overreliance on AI, potential for common regulatory blind spots, and outdated models, are also a concern, leading some authorities to focus on micro-processes for AI implementation.
- Vendor-related risks, including overreliance on a limited number of providers and potential vendor lock-in, are amplified in EMDEs due to fewer local alternatives and a greater reliance on outsourcing compared to advanced economies.
- Establishing clear AI governance and risk management frameworks, ensuring transparency, and maintaining accountability in AI decision-making are top priorities for EMDE authorities, with 93% identifying this as a key governance aspect.
- Authorities are increasingly focused on developing AI governance frameworks that balance innovation with risk mitigation, emphasizing clear decision-making structures, accountability mechanisms, and regular evaluation of AI tools.
- AI-related consumer risks, particularly fraud and scams exacerbated by AI, cybersecurity, and data privacy, are of high concern, though supervisory understanding and responses vary significantly across EMDEs.
- Most EMDE authorities are not yet creating AI-specific regulatory requirements, opting instead to apply existing financial consumer protection frameworks, but some are considering specific disclosures related to AI use in decision-making.
- Coordination and collaboration among domestic authorities, with the private sector, and across borders are deemed essential for effective AI oversight, addressing regulatory gaps, and sharing best practices.
- EMDE authorities recognize the need for international guidance and standard-setting to prevent regulatory arbitrage and harmonize approaches to AI supervision, especially given the cross-border nature of AI technology and service providers.
- Key forward-looking considerations for EMDE authorities include adopting a strategic approach to AI, focusing on foundational IT and data infrastructure, addressing skills gaps, strengthening monitoring capabilities, and fostering robust coordination and collaboration.