AI's rapid adoption across the financial sector exacerbates the risk of operational disruption leading to systemic instability. Regulators have recognized this challenge, and are considering novel solutions to ensure the threats are effectively managed.
What's Happening
The European Systemic Risk Board (ESRB) convened a panel of experts on AI and systemic risks, including a representative from S&P Global Ratings, on Sept. 27, 2024. Discussions encompassed AI models' ability to uncover hidden patterns in complex data, to make faster and more accurate predictions, and to enhance decision-making, including with regards to risk management. The panel also discussed how AI's application in financial services risks unintended consequences, including job displacement, biased-model outcomes, and increased systemic risks.
Why It Matters
Regulators and supervisory bodies will increasingly need to identify, assess, and mitigate systemic risks arising from AI's increasing sophistication and adoption by the financial sector and beyond.
Those risks include (but are not limited to):
- Dependency on a small pool of technology providers--including in chips, cloud services, specialized software, and advanced generative AI models.
- AI-powered trading's potential to accelerate market contagion arising from asset repricing shocks.
- Greater digital interdependency across financial and non-financial corporations.
- Loss of trust among market participants due to cyber incidents, operational failures, and regulatory breaches, which could foster investor panic and disrupt cross-border capital flows.
What Comes Next
Regulators are grappling with the balance between innovation and financial stability. That will require improved AI literacy and training (among regulators and market participants) and better digital infrastructure.
The AI-focused panel identified elements it considers crucial to adequate financial stability supervision amid growing AI use in the financial sector:
Innovation, including the development of agent-based paradigms and causal AI, to prevent and predict potential instability and systemic risk. AI models that can derive causal inference--rather than probabilistic correlations--from datasets are still in their infancy but could clarify portfolio allocation and reveal potential sources of risk contagion. Similarly, agentic paradigms (which can autonomously achieve specific goals, make decisions, and adapt to new information without human intervention) will improve understanding of concurrent systems in financial markets and provide greater visibility of underlying risks. Such developments could improve and accelerate decision making, including by policymakers.
Controls that foster trust in (and support for) complex AI systems. Confidence in AI technologies and their applications will require the development of control systems that: adequately monitor the data that feeds AI models; model output; and identify unintended consequences.
Digital governance and operational resilience that protects against AI-driven systemic risks. Institutions with effective AI governance frameworks and guardrails protecting their connected digital operational resilience are less likely to be a source of systemic risk. They should also benefit from other competitive advantages, such as greater customer trust and reliability.
AI support for real time regulatory monitoring and transparency. Regulators and supervisory bodies' use of AI-powered technology in risk detection and auditing systems could enable real time risk management of financial institutions' AI-models and permit continuous and interconnected safeguards against systemic risks.
Reduced concentration risk in both chip manufacturing and model development. Concentration risk is currently high for 'standard' AI hardware and software but could diminish with the development of so-called small-language models (AI models that can generate content but are smaller in size and complexity compared with large language models), inferencing systems, and new hardware.
Related Research
- Investment And Talent Are The Keys To Unlocking AI's Potential, July 9, 2024
- Artificial Intelligence Insights, Episode 2: Governance Considerations for AI with DI Rifai, Kjell Carlsson and Bruno Bastit, June 13, 2024
- Can generative AI create a productivity boom?, Jan. 10, 2024
- The AI Governance Challenge, Nov. 29, 2023
- AI In Banking: AI Will Be An Incremental Game Changer, Oct. 31, 2023
Other Research
- New Frontiers In Macroprudential Policy, ESRB 8th Annual Conference, Sept. 26-27, 2024
This report does not constitute a rating action.
| Primary Credit Analysts: | Miriam Fernandez, CFA, Madrid + 34917887232; Miriam.Fernandez@spglobal.com |
| Sudeep K Kesh, New York + 1 (212) 438 7982; sudeep.kesh@spglobal.com |
No content (including ratings, credit-related analyses and data, valuations, model, software, or other application or output therefrom) or any part thereof (Content) may be modified, reverse engineered, reproduced, or distributed in any form by any means, or stored in a database or retrieval system, without the prior written permission of Standard & Poor’s Financial Services LLC or its affiliates (collectively, S&P). The Content shall not be used for any unlawful or unauthorized purposes. S&P and any third-party providers, as well as their directors, officers, shareholders, employees, or agents (collectively S&P Parties) do not guarantee the accuracy, completeness, timeliness, or availability of the Content. S&P Parties are not responsible for any errors or omissions (negligent or otherwise), regardless of the cause, for the results obtained from the use of the Content, or for the security or maintenance of any data input by the user. The Content is provided on an “as is” basis. S&P PARTIES DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, ANY WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE OR USE, FREEDOM FROM BUGS, SOFTWARE ERRORS OR DEFECTS, THAT THE CONTENT’S FUNCTIONING WILL BE UNINTERRUPTED, OR THAT THE CONTENT WILL OPERATE WITH ANY SOFTWARE OR HARDWARE CONFIGURATION. In no event shall S&P Parties be liable to any party for any direct, indirect, incidental, exemplary, compensatory, punitive, special or consequential damages, costs, expenses, legal fees, or losses (including, without limitation, lost income or lost profits and opportunity costs or losses caused by negligence) in connection with any use of the Content even if advised of the possibility of such damages.
Credit-related and other analyses, including ratings, and statements in the Content are statements of opinion as of the date they are expressed and not statements of fact. S&P’s opinions, analyses, and rating acknowledgment decisions (described below) are not recommendations to purchase, hold, or sell any securities or to make any investment decisions, and do not address the suitability of any security. S&P assumes no obligation to update the Content following publication in any form or format. The Content should not be relied on and is not a substitute for the skill, judgment, and experience of the user, its management, employees, advisors, and/or clients when making investment and other business decisions. S&P does not act as a fiduciary or an investment advisor except where registered as such. While S&P has obtained information from sources it believes to be reliable, S&P does not perform an audit and undertakes no duty of due diligence or independent verification of any information it receives. Rating-related publications may be published for a variety of reasons that are not necessarily dependent on action by rating committees, including, but not limited to, the publication of a periodic update on a credit rating and related analyses.
To the extent that regulatory authorities allow a rating agency to acknowledge in one jurisdiction a rating issued in another jurisdiction for certain regulatory purposes, S&P reserves the right to assign, withdraw, or suspend such acknowledgement at any time and in its sole discretion. S&P Parties disclaim any duty whatsoever arising out of the assignment, withdrawal, or suspension of an acknowledgment as well as any liability for any damage alleged to have been suffered on account thereof.
S&P keeps certain activities of its business units separate from each other in order to preserve the independence and objectivity of their respective activities. As a result, certain business units of S&P may have information that is not available to other S&P business units. S&P has established policies and procedures to maintain the confidentiality of certain nonpublic information received in connection with each analytical process.
S&P may receive compensation for its ratings and certain analyses, normally from issuers or underwriters of securities or from obligors. S&P reserves the right to disseminate its opinions and analyses. S&P's public ratings and analyses are made available on its Web sites, www.spglobal.com/ratings (free of charge), and www.ratingsdirect.com (subscription), and may be distributed through other means, including via S&P publications and third-party redistributors. Additional information about our ratings fees is available at www.spglobal.com/usratingsfees.