European Central Bank Banking Supervision has published a revised guide to internal models, updating its supervisory expectations for banks’ internal models for credit, market and counterparty credit risk to reflect changes in regulatory requirements under the Capital Requirements Regulation (CRR3) and the revised Basel framework. The revised guide also adds detail on the acceptable use of machine learning techniques and is intended to support transparency and harmonisation in internal model supervision. Key changes include a new section on machine learning in the chapter on overarching principles, setting expectations for explainability and for model performance to justify added complexity. The credit risk chapter aligns roll-out and permanent partial use expectations with CRR3, refines expectations on internal validation and internal audit in line with the European Banking Authority’s supervisory handbook on validation of internal ratings-based systems, and clarifies senior management and management body responsibilities for ensuring readiness of internal model applications submitted to the ECB. It also updates expectations on the definition of default and on estimating probability of default and loss given default. Market risk has been split into separate CRR2 and CRR3 chapters to reflect European Commission decisions delaying implementation of the new Basel standards until the beginning of 2026 and an intended further delay until the beginning of 2027 that is subject to a three-month scrutiny period by the European Parliament and Council, while the counterparty credit risk chapter adds detail on modelling trade-related risks, exposure changes and maturity consistent with CRR3. The revision was developed with experts from national competent authorities and incorporated industry feedback from two targeted stakeholder consultations. ECB Banking Supervision noted that internal models remain a material driver of risk-weighted exposure amounts for significant banks, with around 60% of credit risk risk-weighted exposure amounts calculated using internal models as of June 2024.
European Central Bank - Banking Supervision 2025-07-28
European Central Bank Banking Supervision revises internal models guide to reflect CRR3 and clarify expectations for machine learning
The European Central Bank Banking Supervision updated its guide to internal models, reflecting changes under CRR3 and the revised Basel framework. Key updates include expectations for machine learning, credit risk model validation, and senior management responsibilities, alongside adjustments for market and counterparty credit risk. The revision aims to enhance transparency and harmonisation, with internal models accounting for about 60% of credit risk-weighted exposure for significant banks as of June 2024.