Updates to the supervisory guides on internal models (EGIM, EGAM and EGMA)

The European Central Bank (ECB) has updated its Guide on Internal Models (EGIM). Following the July 2025 revision, the June 2026 version introduces a one-off update by removing the content relating to the credit conversion factor, pending specific guidelines from the European Banking Authority (EBA). At the same time, the ECB updated the Guide on Assessment Methodology and the Guide on Materiality Assessment, removing references to the advanced method for credit valuation adjustment risk following its replacement by the standardized approach.

Updates to the supervisory guides on internal models

Watch video

Executive summary

The revised Guide to Internal Models retains the fundamental principles of model governance, including requirements for documentation, data governance, and consistency at a consolidated level. It also retains the sections on climate and environmental risks, the use of machine learning techniques, and accountability in consolidation and outsourcing processes. Finally, the strengthened supervisory expectations on the use of internal credit, market, and counterparty credit risk models are retained, reflecting the latest regulatory developments, including the Capital Requirements Regulation (CRR) III.

Main content

This Technical Note summarizes the update of the EGIM:

• General principles. The requirements relating to model governance, documentation, data governance, validation, and internal audit remain in place. Similarly, the specific guidance on environmental and climate risks, the use of modified or extended models, and the involvement of third parties is retained. Furthermore, the specific expectations regarding the use of machine learning remain in place, including criteria on explainability, justified complexity, and data standards.
• Credit risk. The criteria for the deployment and permanent partial use of the internal ratings-based (IRB) approach are retained, as are the clear criteria for choosing between the IRB and standardized approaches. Governance, data and validation requirements remain in place, and the definition of default is harmonized. The guidance continues to set out detailed expectations for the estimation of risk parameters such as probability of default (PD) and loss given default (LGD), including requirements for data quality, robustness testing and margins of conservatism (MoC).
• Market risk. The differentiated expectations for the CRR III framework are maintained, with clear guidelines for the transition. The guidance retains the validation, documentation and supervisory requirements for Value-at-Risk (VaR) methodologies, stressed VaR, expected loss, incremental risk charge (IRC), and the management of risks not included in the model engines.
• Counterparty risk. The requirements regarding the risk horizon (MPoR), initial margins, collateral and the granularity of simulations remain in place. The guidance retains the expectations regarding parameter calibration, the treatment of expected effective positive exposure (EEPE), the justification of the alpha parameter and the consideration of risks not captured by the EEPE, whilst also reinforcing the use test and independent validation.

Download the technical note on the Updates to the supervisory guides on internal models (also available in Spanish).