Risk and AI (RAI) Certificate
Model the risk. Govern the model. Explain the decision.
A complete preparation programme for the GARP Risk and AI (RAI) certificate, mapped to the 2026 published learning objectives. It starts from what an algorithm is and ends at model governance, validation, bias, explainability and the controls a regulator expects around automated decisions.
Machine learning · Model risk · Validation · Bias & fairness · Explainability · AI governance · Regulatory expectations
6 modules · 127 chapters
M1 Essential Foundations (start here with zero background) Ch 1–5 · 5 chapters
- AI from zero: how computers learn, decide, and failFoundation
- Basic mathematics for ML: variables, functions, graphs and averagesFoundation
- Statistics foundations: distributions, probability, uncertainty and errorFoundation
- Business decisions, controls and risk: beginner vocabularyFoundation
- Reading AI research claims and model evaluation chartsFoundation
M2 AI and Enterprise Risk: Foundations Ch 6–15 · 10 chapters
- How AI Became an Enterprise Risk Question
- Symbolic AI, Logic, Search and Recursion
- Where Classical AI Breaks Down
- Neural Networks and Representation Learning
- Deep Learning: Breakthroughs and Limitations
- Milestones That Changed AI Adoption
- Four Learning Paradigms Compared
- AI Applications Across Business and Risk
- Black Boxes, Uncertainty and Overreliance
- AI Risk from Individual to Systemic Scale
M3 AI Tools and Techniques Ch 16–53 · 38 chapters
ML workflow, data and preprocessing · Unsupervised learning · Econometric supervised learning · Machine-learning prediction techniques · Semi-supervised learning · Reinforcement learning · Model fitting and optimization · Model-performance evaluation · Natural-language processing · Generative AI and language models
- Machine Learning versus Econometric Modeling
- Understanding and Cleaning Model Data
- Transforming Features and Reducing Dimensions
- Training, Validation, Testing and Leakage
- Clustering and Similarity in Risk Data
- K-Means by Hand: A Worked Example
- Selecting Clusters and Hierarchical Analysis
- Linear Regression for Prediction and Explanation
- Reading Regression Results and Diagnosing Failure
- Logistic Regression versus Discriminant Analysis
- Classification and Regression Trees
- Pruning and Tree Ensembles
- Nearest Neighbors and Support Vector Machines
- Neural Networks, Convolutions and Sequences
- Autoencoders and Model Complexity
- When Labels Are Scarce: Semi-Supervised Learning
- Self-Training versus Co-Training
- Reinforcement Learning: States, Actions and Rewards
- Exploration, Exploitation and Multi-Armed Bandits
- Value Learning, Policy Learning and Deep RL
- Least Squares and Maximum Likelihood
- Optimization by Gradient Descent
- Backpropagation and Training Stability
- Bias, Variance, Underfitting and Overfitting
- Regularization and Resampling for Robust Models
- Hyperparameter Search and Interpretability Tradeoffs
- Model Accuracy Beyond a Single Number
- ROC Curves and Business Decision Thresholds
- The NLP Pipeline and Text Representation
- TF-IDF, Dictionaries and Naive Bayes
- Evaluating Language Classification Systems
- Generative AI: Modalities and Model Families
- Embeddings and Word2Vec Explained
- From RNNs to LSTMs to Transformers
- How Large Language Models Are Trained
- Prompts, Context Windows and Stateless Interaction
- Temperature, Top-K and Top-P
- Evaluating GenAI, Agentic AI and Business Uses
M4 AI Risks, Fairness, Explainability and Safety Ch 54–77 · 24 chapters
- Constructing an AI Risk Taxonomy
- How AI Harm Happens
- Fairness Is Not a Single Formula
- Individual Fairness in Decision Systems
- Group Fairness and Protected Outcomes
- Selection-Rate Parity
- Error-Rate Fairness
- Predictive Parity and Calibration
- The Fairness Tradeoff Problem
- Sampling, Representation and Selection Bias
- Historical, Label and Measurement Bias
- Feedback Loops and Deployment Bias
- Fairness Mitigation and Its Side Effects
- Interpretability, Explainability and Transparency
- Interpretable Models versus Post-Hoc Explanations
- Local and Global Explanations
- When Explanations Mislead
- AI Manipulation and Human Autonomy
- Automation Bias and Human Oversight Failure
- Robustness, Safety and Security Failures
- Human Wellbeing and Social Harm
- Reputation, Liability and Crisis Response
- Systemic, Global and Extreme Risk Scenarios
- Generative AI Risks and Practical Safeguards
M5 Ethics, Privacy and Responsible AI Ch 78–101 · 24 chapters
- Why Organizations Need Applied AI Ethics
- Turning Ethical Intent into Decisions
- Consequentialism in AI Decisions
- Duties and Rights in AI Design
- Virtue Ethics and Professional Judgment
- Choosing Among Ethical Frameworks
- Lessons from Medical Ethics for AI
- Nonmaleficence and Avoidable Harm
- Beneficence and Positive Value
- Justice and Equitable Treatment
- Autonomy, Consent and Human Control
- Explainability as an Ethical Obligation
- Privacy, Confidentiality and Data Minimization
- Consent, Secondary Use and Context
- Cybersecurity as a Responsible-AI Duty
- Ethical Responses to Bias and Discrimination
- Ethics versus Business Incentives
- Accountability and Human-in-the-Loop Design
- Who Must Be Consulted?
- AI Ethics Principles in Organizations
- AI Regulation: Understanding Types of Obligations
- Comparing Global AI Governance Approaches
- Ethical Impact Assessments and Assurance
- Ethics Board Simulation: A High-Stakes AI Decision
M6 Data and AI Model Governance Ch 102–127 · 26 chapters
- Why Data Governance Determines AI Reliability
- Data Owners, Stewards and Control Roles
- Data Quality, Provenance and Traceability
- Governing Data Across Its Lifecycle
- Enterprise Model Governance Architecture
- AI Model Classification and Risk Tiering
- Creating and Maintaining a Model Inventory
- Three Lines of Responsibility for Model Risk
- From Business Need to Model Specification
- Model Documentation That Supports Challenge
- Development Testing and Release Evidence
- Implementation, Integration and Change Readiness
- Why Independent Model Validation Matters
- Conceptual Soundness and Developmental Evidence
- Outcomes Analysis, Benchmarking and Backtesting
- Ongoing Monitoring, Drift and Degradation
- Model Risk Policies, Limits and Exceptions
- Change Control, Retraining and Versioning
- Periodic Review and Model Challenge
- Retirement, Archival and Post-Use Risk
- How Model Results Are Misinterpreted
- Governing Generative AI Data and Outputs
- Third-Party and Agentic AI Governance
- An Enterprise AI Model Risk Committee Case
- AI Governance Boardroom Capstone (Original Case)Optional
- Practical No-Code Model DemoOptional
Three ways to take Risk and AI (RAI) Prep Programme
| Feature | Self-paced | Cohort | Enterprise |
|---|---|---|---|
| Format | Written chapters, video explainers and podcasts | Everything in self-paced, plus scheduled live sessions | Everything in cohort, delivered privately to your team |
| Live sessions | None | Scheduled, instructor-led | Scheduled, instructor-led, private |
| Mentorship | Not offered | Not offered | Not offered |
| Access | Lifetime | Lifetime | Lifetime for every enrolled seat |
| Pace | Entirely your own | Guided schedule with a peer group | Agreed with your desk |
| Tailoring | Fixed curriculum | Fixed curriculum | Sequenced to your markets, systems and governance |
| Best for | Individuals learning around a job | Individuals who want structure and deadlines | Desks building the same capability together |
Get the programme guide
The full chapter list, what each module covers, and how the tiers compare - sent to your inbox as a PDF.
Run this for my desk
Private delivery for your desk, sequenced to your markets and systems. Tell us the team and we will scope it.
Run this for my deskBring this to your desk
Download the full curriculum, or talk to us about running it for a team.
FRM, SCR and Risk and AI (RAI) are trademarks of the Global Association of Risk Professionals (GARP). Durga Analytics is an independent training provider. It is not affiliated with, endorsed by or a partner of GARP, and it does not award these credentials. GARP does not endorse, promote or warrant the accuracy of this material. Exam scope, weightings and dates are set by GARP: verify them against the official candidate guide before you sit.