Banks and financial institutions have been using models for decades, for example to manage risks or to calculate prices. Hence, there has always been the risk that an incorrect model specification or the incorrect use of a model lead to a decision coming along with negative consequences, such as financial losses. This risk is also referred to as Model Risk.
To measure and mitigate Model Risk, banks have established extensive and complex Model Risk Management (MRM) approaches. However, with the increasing use of Artificial Intelligence (AI) and Machine Learning (ML), a comprehensive adaptation of this Model Risk approach becomes essential.
Which risks and regulatory requirements need to be considered
Despite a wide range of undisputed benefits, AI/ML approaches come along with their own specific inherent risks that are not yet adequately addressed in the existing Model Risk Management frameworks. In the future, it is expected that AI/ML models will also be increasingly used in areas which were previously unsupported or poorly supported by models. In addition, there are growing regulatory requirements such as the "EU Artificial Intelligence Act".
Seven key pillars for the use of AI in Risk Modeling
In order to meet the new challenges posed by AI/ML models and the corresponding regulatory requirements, it is advisable to enhance or extend existing MRM approaches in specific areas. These enhancements can be integrated into already existing frameworks, processes and IT-tools, thus generating synergies.
In the white paper "Modern Risk Management for AI Models", written in cooperation with KPMG India, we derive seven key pillars for the enhancement of the Model Risk Management framework based on the specific requirements of AI/ML algorithms.