OSFI Guideline E-23

Model Risk Management: Overview and Considerations

An independent overview of OSFI’s updated guidance and its relevance to model data, automated valuation models (AVMs) and third-party model use. Prepared as an educational resource for Canadian financial institutions.Effective May 1, 2027

E-23 at a glance

May 1, 2027

The effective date of OSFI’s final updated Model Risk Management guideline, published September 11, 2025.

Federally regulated institutions

The guideline applies to federally regulated financial institutions, subject to its scope and application provisions.

Risk-proportionate governance

Model-risk management expectations are applied according to the institution’s circumstances and the risk of a model.

What is OSFI Guideline E-23?

The Office of the Superintendent of Financial Institutions (OSFI) issued the final Guideline E-23 – Model Risk Management (2027) on September 11, 2025. It sets out expectations for effective enterprise-wide management of risks arising from models used by federally regulated financial institutions (FRFIs).

The guideline includes models using artificial intelligence and machine learning, and covers relevant models and data obtained from external sources, including third-party vendors. Its application is risk based: not every model is subject to the same intensity of governance.

E-23 is guidance directed at regulated institutions. It does not certify technology vendors, mandate a particular property dataset or designate any AVM as compliant.

Selected model-risk management considerations

E-23 considers the model lifecycle, including model rationale, design, review, deployment and ongoing monitoring. For property-related analytical models, three relevant areas are:

Model data

Principle 3.2 addresses whether data used to develop a model are suitable for the intended use, including accuracy, relevance, representativeness and data quality.

Independent model review

Principle 3.4 addresses independent assessment of conceptual soundness, performance and fitness for purpose, with review proportionate to risk.

Deployment and monitoring

Principles 3.5 and 3.6 address quality and change control during deployment, ongoing monitoring and model decommissioning.

Where property data and AVMs fit

Automated valuation models generate property-value estimates from available data and modelling methods. Their suitability depends on the institution’s intended use, the underlying evidence, model limitations and relevant oversight. An AVM estimate is not a guaranteed sale price or a substitute for every form of professional appraisal.

Property attributes and input quality

Living area, property type, location and other characteristics may inform a valuation model. Source-record values, derived building footprints and AI-estimated characteristics should be distinguished when assessing quality and uncertainty. Completeness can vary by region, property type and source.

Third-party models and data

Institutions remain responsible for assessing model risk when obtaining models or data from external providers. E-23 references OSFI’s Guideline B-10 for applicable third-party risk management principles. Vendor sourcing does not replace institution-led governance.

Property intelligence from Gnowise

Gnowise offers Canadian residential property data, automated valuation models and related analytics for institutional and professional use, subject to available product scope and geographic coverage. Institutions can review the relevance and limitations of these services for their own use cases.

Property Attributes

Explore available property characteristics, modelled living-area estimates, geospatial information and bulk data options.

Intelligent Valuation Models

Learn about Canadian property-value estimates, confidence indicators and relevant model limitations.

WISE1 API

Explore the integration of available property intelligence into lender and other institutional workflows.

Gnowise does not determine an institution’s regulatory obligations, certify E-23 compliance or replace independent model review. OSFI’s official guideline remains the authoritative source.

FAQs

OSFI states that the final guideline takes effect on May 1, 2027 for federally regulated financial institutions.

No. The guideline takes a risk-based approach. Institutions identify models and assess inherent model risk; the intensity of model lifecycle governance is proportionate to model risk and the institution’s circumstances.

No. E-23 sets model-risk management expectations. It does not prescribe a particular property dataset, AVM vendor or universally required set of real-estate fields.

No. Institutions remain responsible for evaluating and managing risks of externally sourced models and data. Appropriate review, monitoring and governance depend on a model’s role and risk.

Official sources and related reading

About this resource. General educational information prepared by Gnowise and reviewed November 2026. It is not legal or regulatory advice, an assertion of product compliance or an OSFI endorsement. Refer to the official guideline and your own regulatory and model-risk specialists for institution-specific interpretation.