The Global Risks Alliance (GRA)
Model risk is the potential for loss or harm from decisions based on models that are wrong or misused. Banks, insurers and asset managers manage it through model inventories, risk tiering, independent validation, approval and ongoing monitoring, overseen by boards and senior management. Machine learning and generative AI widen what counts as a model. This page sets out how model risk governance works and fails, the records involved, and who decides.
GRA · In development · service level R0
Model risk governance is the set of policies, roles and controls through which financial firms identify, validate, approve, monitor and retire models, and boards oversee model risk. Machine learning and generative AI widen its scope. Boards, model risk functions, supervisors and internal audit hold the decisions.
How satellite radar turns a flood into an exposure update. Synthetic scene; no real event or place.
The working screen behind this subject: its structure, terms and states. Names are placeholders and values illustrative.
Model risk governance · Model inventory · Committee 9
| MD-204 PD retail | Tier 1 | 14 months | 3 | Overdue |
| MD-511 LLM complaints triage | Tier 1 | never | 0 | Not validated |
| MD-164 IFRS 9 staging | Tier 1 | 8 months | 1 | Approved |
| MD-534 marketing propensity | Tier 3 | 17 months | 0 | Approved |
Inventory, ownership and validation. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.
Spreadsheets, vendor tools and AI components often sit outside the model inventory. Unlisted models escape validation and monitoring, and their errors surface only when losses or complaints arrive.
Validation that lacks independence, skill or authority to stop a model becomes a formality. Effective challenge needs validators who understand the method and can block use.
Models are tiered by materiality and complexity, which decides validation depth. Underestimating a model's materiality, or the reach of an AI tool, leaves high-impact models lightly reviewed.
Outputs of one model feed others: a score feeds pricing, which feeds capital. An error upstream propagates, and aggregate model risk across the chain is rarely measured.
Each term as this page uses it. Where a standard-setting body defines the term, the wording follows that body.
| Term | Definition |
|---|---|
| Model tiering | Classifying models by materiality and complexity to set how deeply each is validated and monitored. |
| Model owner | The person accountable for a model's development, use and performance within a firm. |
| Model use restriction | A limit placed on how a model may be used until identified weaknesses are fixed. |
Each indicator carries its unit, its cadence, the kind of source behind it and the evidence label a value would carry. No values are shown: a programme supplies them under its own authority.
| Indicator | Unit | Cadence | Source type | Label |
|---|---|---|---|---|
| Models in inventory | count by tier | quarterly | Model inventory | DECLARED |
| Models validated within policy cycle | share by tier | quarterly | Validation records | DECLARED |
| Open validation findings | count by severity | quarterly | Validation records | DECLARED |
| Models used with exceptions or restrictions | count | quarterly | Model governance records | DECLARED |
| AI and machine learning models in inventory | count and share | annual | Inventory and surveys | DECLARED |
| Supervisory findings on model risk management | count | annual | Supervisory reports | OFFICIAL |
The topic as a computational graph: named sources feed the methods, an evidence review gates the outputs, and the decisions stay with the authorities that hold them.
Fig. 02Nodes carry their type, two metric rows and a state; connectors show what feeds what; dashed frames name the owners; the timeline is the cadence. Design of the pipeline, not a record of a run.
Named public datasets and institutional records, each with its licence class. Sources by agreement are used only under their terms and never republished.
| Source | What it gives | Licence class |
|---|---|---|
| Federal Reserve and OCC SR 11-7 guidance | Model risk management expectations | Public domain |
| PRA SS1/23 model risk management principles | UK principles for banks | Free, Bank of England terms |
| ECB guide to internal models | Expectations for internal models used for capital | Free, ECB terms |
| Bank of England and FCA surveys of AI and machine learning in UK financial services | Use and governance of AI across firms | Free, Bank of England and FCA terms |
| EBA follow-up reports on machine learning for internal ratings-based models | Supervisory views on ML in capital models | Free, EU reuse terms |
AI and Financial Model Risk prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.
| Who decides | The question | Evidence this page serves |
|---|---|---|
| Boards and senior management | What model risk appetite to set and how to oversee it | Inventory, tiering, findings and incident evidence |
| Model risk management functions | Which models to validate, approve, restrict or retire | Validation results and monitoring |
| Prudential supervisors | Whether model risk management meets expectations | Supervisory reviews and firm reporting |
| Internal audit | Whether the model risk framework operates as designed | Governance and validation records |
Governance questions begin with knowing what models exist.
Models fail through flawed design, poor data, implementation errors, or use beyond their intended scope. Governance frameworks address these through development standards, independent validation and ongoing monitoring, scaled to each model's materiality. Supervisory guidance on model risk management in several jurisdictions sets expectations along these lines.
AI strains traditional frameworks. Machine learning models can be numerous, frequently retrained and embedded in processes; generative AI tools may be adopted informally by staff; vendor models may be opaque. Inventories that miss these uses leave risk unmanaged.
| Method or data | What it shows | Limits |
|---|---|---|
| Model inventories and tiering | Which models exist and how material they are | Informal and embedded AI uses are easily missed |
| Independent validation | Soundness of design, data and implementation | Validation is constrained by time, skills and access to vendor details |
| Governance benchmarking | How institutions organize model risk | Self-reported and hard to compare |
| Issue and finding analysis | Recurring weaknesses | Findings are confidential within firms |
The current service level is R0: information and enquiries. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.