Models & quantification

Model risk

Model risk is the risk of loss from decisions based principally on the output of models, caused by errors in the design, development, parameter estimation, implementation, use or monitoring of those models. For banks this definition has been part of the CRR since 1 January 2025, where model risk falls under operational risk. The concept applies to any model on which an organisation bases its decisions.

Source: Regulation (EU) No 575/2013 (CRR) Article 4(1), points (52) and (52b), as amended by Regulation (EU) 2024/1623 (CRR3); Directive 2013/36/EU (CRD) Article 3(1), point (11); ECB Guide to internal models (June 2026), Overarching principles, section 4; Directive 2009/138/EC (Solvency II) Article 120

Banks, insurers and pension funds base decisions on matters such as lending, pricing, provisions, capital and liquidity on models. Model risk arises when a model is wrong or is used wrongly. The CRR gives examples: flawed design, inadequate verification of whether the model suits the product and the market conditions, implementation errors, use for a purpose the model was not designed for, including manipulation of parameters, and untimely or ineffective monitoring or validation of its performance.

The definition is in Article 4(1), point (52b), of Regulation (EU) No 575/2013 (CRR), inserted by CRR3. Since CRD VI, Article 3(1), point (11), of Directive 2013/36/EU (CRD) also refers to it; documents from before 2026 still use the older CRD wording on errors in development, implementation or use. The ECB Guide to internal models (June 2026, non-binding) expects banks to have a framework for managing model risk, with a written policy that sets out what counts as a model, a register of models, methods to assess and measure model risk, and clear roles for development, validation and internal audit. For insurers using an internal model, the administrative, management or supervisory body is responsible for ensuring that the design and operation of the model remain appropriate (Article 120 of Directive 2009/138/EC, Solvency II).

Model risk is part of operational risk (CRR Article 4(1), point (52)) and differs from the risks a model measures, such as credit risk or market risk. The main controls are a complete model inventory and independent model validation. Questions for the supervisory board and the audit committee: which decisions rely on which models, which models are running with open validation findings, and does the model policy also cover artificial intelligence? For AI systems that assess the creditworthiness of natural persons, the high-risk obligations of the AI Act apply from 2 December 2027.

Last updated: · Editor: ONE Risk Advisory

Frequently asked questions

Do the US SR 11-7 and SR 26-2 apply to Dutch banks?

No. SR 11-7 from 2011 and its successor SR 26-2 of 17 April 2026 are supervisory guidance from the Federal Reserve, the FDIC and the OCC for US banking organisations. Dutch banks are subject to the definition in the CRR, the reference in the CRD and the expectations in the ECB Guide to internal models. The US document can still serve as a point of comparison.

Is model risk only a concern for banks?

No. The legal definition sits in the banking rules, but insurers with an internal model must also validate that model regularly, and their administrative, management or supervisory body remains responsible for its design and operation (Solvency II, Articles 120 and 124). Other organisations can apply the concept to any model that drives prices, provisions or forecasts, even where no supervisor requires it.

← All terms