FSB Examines AI Adoption Across the Financial Sector

In a report, the FSB said that its work builds on a 2024 AI study and responds to a request from the South African G20 Presidency to explore how regulators can strengthen oversight of AI adoption and manage systemic risks. Drawing from member surveys, regulator interviews, and public data, the FSB evaluated current supervisory practices, identified key definitional and data gaps, and proposed indicators to enhance consistency in AI monitoring across jurisdictions.

The FSB structured its report around the following:

1. Early Stages of AI Oversight. The FSB said that financial authorities are still in the early stages of developing frameworks to monitor AI adoption and related risks. The FSB reported that current oversight relies on surveys, supervisory engagement, and analysis of public data. The FSB explained that these efforts are limited by inconsistent definitions, lack of cross-border comparability, and the high cost of gathering reliable information. The FSB noted that difficulties in assessing the "criticality" of AI systems have further constrained regulators’ ability to evaluate potential financial stability implications.
2. Development of Monitoring Indicators. The FSB proposed a framework of direct and proxy indicators to help authorities monitor AI adoption and identify related vulnerabilities. The FSB stated that these indicators should capture exposure across key risk areas such as AI adoption, third-party dependencies, market correlations, cyber threats, and model governance. The FSB emphasized that monitoring should be proportionate, risk-based, and harmonized across jurisdictions to ensure consistency and reduce reporting burdens.
3. Case Study on Generative AI. The FSB reviewed generative AI ("GenAI") as a case study, highlighting concentration and dependency risks within its layered supply chain of hardware, cloud infrastructure, and pre-trained models. The FSB found that financial institutions are increasingly dependent on a small number of major technology providers for GenAI capabilities, as developing these systems internally is often cost-prohibitive. The FSB warned that this reliance creates potential single points of failure and systemic vulnerabilities and urged regulators to apply its "Third-Party Risk Management Toolkit" to evaluate the criticality, substitutability, and systemic importance of GenAI services.
4. Recommendations for Regulators. The FSB recommended that: (i) national authorities should strengthen monitoring frameworks by adopting the indicators outlined in the report; (ii) supervisors should enhance coordination and data sharing across domestic and international regulators to align taxonomies and best practices; and (iii) the FSB and standard-setting bodies should continue monitoring AI developments and address data gaps in areas such as market correlations, model governance, and misaligned AI systems.