OSFI Guideline E-23 · Model Risk Management

AI Model Risk Alignment Tool

For AI and Machine Learning Models at Federally Regulated Financial Institutions

Tier an AI or machine learning model against OSFI Guideline E-23 in about five minutes. Free, with no sign-up. It runs entirely in your browser: nothing you type is sent to a server, stored by us, or seen by anyone.

The 7 risk dimensions it asks about

  • Financial & Member Impact — If this model produces materially incorrect outputs for one week, what is the estimated impact? (E-23 §3.2(a))
  • Decision Breadth & Reach — How many member decisions or transactions does this model influence per month? (E-23 §3.2(b))
  • Model Complexity — How would you characterize the technical complexity and interpretability of this model? (E-23 §3.2(c))
  • Degree of Automation — How much human oversight exists before model outputs affect members or operations? (E-23 §3.2(d) & HGAI Principle 4)
  • Data Sensitivity — What categories of data does this model consume? (E-23 §3.2(e) & PIPEDA)
  • Third-Party Dependency — Does this model rely on external vendor components or APIs? (E-23 §8)
  • Business Criticality — What is the consequence if this model becomes unavailable? (E-23 §3.2(f))

What you get

  • A materiality tier from 1 (high) to 3 (low), and the rule that produced it, so the number can be defended to a second line or an examiner.
  • Which of the 10 E-23 governance controls that tier requires, and which you already have.
  • For generative AI, retrieval and multi-agent systems, 6 AI-specific controls on top.
  • A printable report, and a model-inventory.yaml block for osfi-gate to enforce.