AI model risk: What the current model risk management framework can teach us about managing the risks of AI models

Abstract

The rapid adoption of Artificial Intelligence (AI) among financial institutions in recent years creates several opportunities, but also presents significant risks that require adequate risk management. Despite advances in recent years, AI regulation remains fragmented. This creates a challenge for financial institutions when looking for guidance on how to address the emerging risks presented by the use of AI. Given the complexity and speed of revision, AI models tend to propagate and amplify existing model risk. This grants them the potential to be more harmful, and raises important model ethics concerns. This paper discusses how the existing model risk management framework can offer important lessons for financial institutions on how to tackle these emerging risks. Additionally, the paper explores possible enhancements to the model risk management framework in order to address the unique challenges posed by AI models. These include adapting governance and policies, including model ethics considerations; enhancing model risk identification and classification; and updating model life cycles, with an emphasis on data management, model development, validation and monitoring. While the author agrees that AI risks are diverse in nature, the focus of the paper is on the risks derived from the use and development of AI models.