THE SIGNAL
On September 11, 2025, the Office of the Superintendent of Financial Institutions (OSFI) published final Guideline E-23, Model Risk Management. It takes effect May 1, 2027 and applies to all federally regulated financial institutions, including AI and machine learning models and models sourced from vendors and third parties.
Two details matter for speed. E-23 defines a Model Approver, the person or committee with authority to approve a model’s use. And it says the inherent model risk rating should drive the “level of authority required to approve the model.” Approval is meant to be proportional to risk.
BEFORE
Hypothetical: a Canadian lender’s credit team recalibrates a small-business adjudication model. Independent review has signed off with two minor findings and compensating controls documented. Because every model change goes to the monthly model risk committee, the update waits three weeks.
Loan officers keep using the old version, and overrides pile up. The file is ready. The person who can release it is not.
WHAT THE WAIT COSTS
A stale model keeps making decisions. Overrides create their own risk and their own paperwork. The committee spends its hours on low-rated changes, which leaves less attention for the high-rated models that deserve it.
AFTER
Approval authority follows the risk rating. A low-rated change with a clean review goes to a named delegate with the reviewer’s findings and the inventory entry attached, and it is approved within days.
The committee spends its time on the high-rated models. Every approval is recorded against the model ID, which is the trail a supervisor will ask to see.
HOW WISDOMTWIN FITS
WisdomTwin.ai is building a sovereign judgment layer for regulated enterprises in banking, healthcare, legal, government, pharma and defense. It is designed to run inside your own environment, assemble the decision packet with the evidence attached, route it to the named person who holds the authority, and keep a record of who released the decision and why.
We don’t replace judgment. We remove the wait.
TRY THIS THIS WEEK
Pull your last ten model approvals. Note each risk rating and the days from review sign-off to approval.
If low-rated and high-rated changes waited about the same time, your approval authority is not yet risk-based.
Draft a one-page delegation table: risk rating, approver, maximum wait.
SOURCES
OSFI, Guideline E-23, Model Risk Management (2027), September 11, 2025: https://www.osfi-bsif.gc.ca/en/guidance/guidance-library/guideline-e-23-model-risk-management-2027
Examples in this issue are hypothetical. WisdomTwin demonstrations use synthetic data, and nothing here describes a customer result.
Book a 20-minute call: https://calendly.com/romanbodnarchuk/20min
Roman Bodnarchuk, Co-Founder and CEO, WisdomTwin.ai