The Dutch Authority for the Financial Markets (AFM) published a report on model risk management by asset managers, highlighting that the growing use of complex models and artificial intelligence in portfolio decisions and risk management increases the risk of errors that can lead to poor decisions and financial harm. The report describes current market practices and provides practical guidance to strengthen control of model risk in line with the requirement under the Dutch Financial Supervision Act to maintain controlled and sound business operations. The findings draw on a broad survey of 250 market participants and follow-up questions to 14 asset managers. The AFM sets out eight good practices, including clear assignment of model responsibilities, a shared definition of “model”, explicit risk appetite for model risk, a central model inventory, pre-use and ongoing validation, lifecycle management from development through decommissioning, controls for third-party models, and sufficient in-house expertise to identify and manage risks. The review also notes that while firms recognise the importance of model risk management, definitions and approaches differ widely, governance is often not explicitly documented, and model-risk appetite is seldom formally articulated. The AFM indicated it will continue to monitor the topic and encouraged ongoing dialogue and knowledge sharing with supervisors and model providers, particularly for external models and AI applications.
Dutch Authority for the Financial Markets 2025-12-01
Dutch Authority for the Financial Markets publishes report setting out good practices for asset managers’ model risk management
The Dutch Authority for the Financial Markets (AFM) released a report on model risk management by asset managers, highlighting increased error risks from complex models and AI in portfolio decisions. It outlines eight good practices, including clear model responsibilities, a central model inventory, and sufficient in-house expertise. The AFM will continue monitoring and encourages dialogue with supervisors and model providers, especially regarding external models and AI.