Machine Learning Model Risk and Stress Testing Course
Tier, validate and stress test learned credit and market models with challengers and reverse stress, for model risk and treasury professionals.
This 1-week course has 25 dates from November 2026 to November 2027 in 24 cities and live online. The next session runs 22 – 26 November 2026 in Istanbul. Fees per participant run from 4,000 £ (Cairo) to 9,500 £ (New York) in a classroom, and from 2,900 £ live online.
See dates and book Download brochure (PDF) Training for your team
- Duration
- 1 Week
- Next start
- Istanbul
- Scheduled dates
- 25
- Classroom cities
- 24
- Classroom fee from
- 4,000 £
- Online fee from
- 2,900 £
When and where does this course run?
Showing 1–20 of 25 dates
November 2026
January 2027
February 2027
March 2027
April 2027
May 2027
June 2027
July 2027
August 2027
September 2027
Can’t find a suitable date? Tell us what you need.
What does this course cover?
Machine Learning Model Risk and Stress Testing is a 1-week training course for model risk, validation and treasury risk professionals in financial institutions who challenge learned models, leaving with a Model Stress Validation Report. Gradient boosting and neural networks now sit inside credit, pricing and market risk engines, yet their behaviour under severe scenarios is rarely tested. Jupiter Prime Training runs this course for second-line model risk teams.
Why Machine Learning Model Risk and Stress Testing Matters Now
Three forces converge on model risk functions. Business lines deploy machine learning models faster than validation teams can review them; supervisors and boards expect stress results that cover model behaviour as well as portfolio losses; and recent market shocks exposed models trained only on calm periods. Institutions that tier their models, maintain challengers and stress learned models under severe but plausible scenarios avoid capital surprises and mispriced risk. Those that do not discover non-linear failures only when losses appear. The course draws on ISO/IEC 42001, the international standard for an AI management system, and the NIST AI Risk Management Framework (NIST AI RMF).
Who Gains Most from Machine Learning Model Risk and Stress Testing
- Model risk managers who maintain the inventory and tiering scheme
- Independent validators testing conceptual soundness and performance
- Treasury and market risk staff running scenario and sensitivity tests
- Credit risk modellers building probability of default and loss models
- Internal audit staff reviewing the model lifecycle
A bank or insurer gains most when its model risk manager, validator and stress testing lead attend together.
Machine Learning Model Risk and Stress Testing Capabilities
Participants extend familiar validation discipline to models whose logic is learned rather than specified.
- Assess machine learning models for inventory entry and risk tiering — model risk committee
- Test conceptual soundness, stability and explainability of learned models — independent validation
- Build challenger models to benchmark production performance — model owners
- Design severe but plausible scenarios for AI-driven credit and market models — treasury and risk leadership
- Calculate sensitivity and reverse stress results on model inputs — capital planning
- Communicate validation findings and model limitations — board risk committee
Key Machine Learning Model Risk and Stress Testing Terms
- Model tiering — ranking models by materiality and complexity to set the depth and frequency of validation.
- Challenger model — an alternative model run alongside production to test whether its results remain credible.
- Conceptual soundness — evidence that a model's design, data and assumptions suit its intended use.
- Reverse stress test — working backwards from an unacceptable loss to find the scenarios that would cause it.
- Feature sensitivity — how strongly a model's output moves when one input variable is shocked.
- Out-of-time validation — testing a model on a later period than its training data to expose regime changes.
Machine Learning Model Risk and Stress Testing Day by Day
Day 1: Model Risk in a Learned-Model World
- Model Inventory Extension for Machine Learning — Capturing Hidden Models
- Materiality and Complexity Tiering Matrix — Setting Validation Depth
- Model Lifecycle Map — Development, Approval, Monitoring and Retirement
- ISO/IEC 42001 Management System Links — Where Model Risk Fits
- Banking and Insurance Case Comparison — Credit Versus Reserving Models
Day 2: Independent Validation of Learned Models
- Conceptual Soundness Review — Data, Features and Algorithm Choice
- Out-of-Time and Out-of-Sample Testing — Exposing Regime Dependence
- Explainability Methods for Validators — Feature Attribution and Partial Dependence
- Stability and Population Shift Indices — Detecting Input Drift
- NIST AI RMF Measure Function — Structuring Validation Evidence
Day 3: Challenger Models and Benchmarking
- Challenger Model Design — Simpler Benchmarks Versus Alternative Algorithms
- Champion–Challenger Comparison Metrics — Discrimination, Calibration and Stability
- Override and Overlay Tracking — When Judgement Replaces Model Output
- Vendor Model Validation Limits — Testing What Cannot Be Opened
- Validation Finding Severity Scale — Remediation Deadlines and Escalation
Day 4: Stress Testing AI-Driven Credit and Market Models
- Scenario Design for Learned Models — Severe but Plausible Input Paths
- Feature Sensitivity Shocks — Single-Factor and Joint Stresses
- Reverse Stress Testing — Finding Breaking Points in Model Behaviour
- Credit Loss and Market Value Translation — From Model Output to Capital
- NIST AI RMF Manage Function — Limits, Fallbacks and Model Restrictions
Day 5: Practising the Model Stress Validation Report
- Retail Credit Scoring Case — Applying Tiering, Validation and Challengers
- Trading Book Pricing Case — Applying Sensitivity and Reverse Stress
- Board Risk Committee Briefing — Presenting Findings and Limitations
- Ongoing Monitoring Plan — Thresholds Triggering Revalidation
- Model Stress Validation Report — Evidence, Findings and Use Conditions
Applying Machine Learning Model Risk and Stress Testing After the Course
- Run a tiering review of every machine learning model in the inventory and record any that lack independent validation.
- Lead a reverse stress test on one AI-driven credit model and report the scenarios that would breach risk appetite.
Machine Learning Model Risk and Stress Testing Questions (FAQ)
What is machine learning model risk and stress testing and why does it matter now?
It is the practice of inventorying, validating and stress testing learned models used in financial decisions. It matters now because these models drive credit and pricing outcomes yet can fail non-linearly under scenarios absent from their training data.
Which roles gain most from machine learning model risk and stress testing?
Model risk managers, independent validators, treasury risk staff and credit modellers gain most. You benefit if you approve, challenge or stress models in a bank or insurer.
How does machine learning model risk and stress testing differ from a general model validation course?
It focuses on learned models and joins validation with scenario stress testing. You test feature sensitivity, challengers and reverse stress rather than only statistical fit.
What do you bring back from machine learning model risk and stress testing?
You bring back a validation and stress report template covering tiering, soundness tests, challenger results, scenario outcomes and conditions of use.
Machine Learning Model Risk and Stress Testing: The Way Forward
The Model Stress Validation Report first informs whether a learned model may stay in production unchanged, with restrictions or not at all. Validators, model owners and the risk committee use it at each review. After its first use, revisit the scenario set, because new market conditions often reveal input combinations the original stresses did not cover.
Find more AI Courses for Managers and Professionals by city
Questions about dates, fees and booking
When does this course next run?
The next session runs 22 – 26 November 2026 in Istanbul. It is one of 25 dates scheduled up to November 2027.
Where is this course held?
In a classroom in 24 cities: Istanbul, Barcelona, Abu Dhabi, Paris, Geneva, Dar es Salaam, Doha, London, and more. It is also scheduled as live online sessions.
Can I attend this course online?
Yes. One date is scheduled as a live online session: 10 – 14 January 2027, from 2,900 £ per participant.
How long is the course?
1 Week. Each date shows its start and end dates.
How much does the course cost?
Fees per participant run from 4,000 £ (Cairo) to 9,500 £ (New York) in a classroom, and from 2,900 £ live online. Each date shows its own fee.
Is there a brochure for this course?
How do I book a place?
Choose a date above and select Register, or send us an enquiry.
Can this course be run for our team?
Yes. Tell us the number of participants and your preferred dates and city through our training request form.