ABOUT: Bank of England’s latest publication delves into the evolving landscape of AI and ML in financial services.
LINK: FS2/23 – Artificial Intelligence and Machine Learning
Summary
The Bank of England, in collaboration with the Prudential Regulation Authority and the Financial Conduct Authority, has released a feedback statement on the discussion paper concerning Artificial Intelligence (AI) and Machine Learning (ML) in the financial sector. The document, titled “FS2/23 – Artificial Intelligence and Machine Learning,” is a culmination of responses from various stakeholders to the initial discussion paper published in October 2022. It aims to summarize the feedback without offering policy proposals or indicating future regulatory actions. The feedback underscores the rapid evolution of AI capabilities and the need for ‘live’ regulatory guidance, emphasizing the importance of ongoing industry engagement and calling for greater regulatory coordination both domestically and internationally.
Key Points and Ideas
- A regulatory definition of AI is deemed unnecessary by industry stakeholders.
- There is a call for ‘live’ regulatory guidance that keeps pace with AI advancements.
- Ongoing public-private engagement is crucial for effective AI regulation.
- The complexity and fragmentation of the current regulatory landscape need addressing.
- Data regulation requires more alignment to manage risks related to fairness and bias.
- Consumer outcomes, particularly fairness and ethics, should be central to AI regulation.
- The increasing use of third-party models and data raises concerns that need regulatory guidance.
- A holistic approach across business units is necessary to mitigate AI risks.
- Existing governance structures are considered sufficient to address AI risks.
- Collaboration between data management and model risk management teams is beneficial.
- The principles proposed in CP6/22 are adequate for AI model risk but could be enhanced.
- The Senior Managers and Certification Regime (SM&CR) is seen as adequate for AI risk management.
Key Statistics
- The discussion paper (DP5/22) received 54 responses from a diverse range of stakeholders.
- Industry bodies represented almost a quarter of the respondents, with banks making up a further fifth.
- There was no significant divergence of opinion between different sectors.
Key Take Aways
- The financial industry favors a principles-based or risk-based approach over a fixed regulatory definition of AI.
- Regulatory guidance should be dynamic and periodically updated to reflect the rapid changes in AI technology.
- There is a consensus on the need for more coordinated regulatory efforts at both national and international levels.
- Data-related regulations should be harmonized to effectively address the risks associated with AI, such as bias and fairness.
- The focus of AI regulation and supervision should be on ensuring positive consumer outcomes and addressing ethical considerations.
- The role of third-party AI models and data in the financial sector is expanding, highlighting the need for clearer regulatory guidance.
- A comprehensive approach involving various firm functions is key to managing AI risks effectively.
- Current firm governance structures are generally seen as robust enough to manage AI risks, but there may be areas that require further clarification or strengthening.
RO-AR insider newsletter
Receive notifications of new RO-AR content notifications: Also subscribe here - unsubscribe anytime
