Executive Summary

The Prudential Regulation Authority (PRA) had set out its expectations for model-risk management (MRM) at banks through a consultation paper (CP6/22) published last year.

The paper specifically mentioned the risks associated with artificial intelligence (AI)/ machine learning (ML) models and the need to bring such models within the MRM scope.

The PRA received 11 written responses on AI/ML models in addition to responses around algorithmic trading, particularly dynamic recalibration. The responses indicated keen interest among firms on the topics of AI/ML and algorithmic trading, especially with respect to managing risks associated with them.

The PRA provided feedback on the responses through a policy statement (PS6/23) and recently published its final policy supervisory statement 1/23 (SS1/23), laying down model-risk management principles for banks.

In this report, we present our point of view on the impact of SS1/23 on algorithmic trading and AI/ML.

We endeavour to articulate the key requirements for each of the five principles of MRM SS1/23 for algorithmic trading and AI/ML with a view to support the initial gap analysis and the subsequent self-assessment firms need to carry out in line with the PRA’s guidance.

We also reflect on some of the discussion points from previous forums and round tables in the section ‘Key considerations for initial self-assessment’.

Finally, we seek to highlight some of our expertise in model risk and how we can help firms given our experience in implementing similar regulations such as SR11-7.