AI audit for experienced auditors, covering AI governance, operations and auditing tools and techniques. $459 for members, $599 otherwise, plus a $50 application fee.
Who it's for: IT and financial auditors who already hold CISA or an equivalent.
Eligibility: you must hold an active CISA, or a qualifying audit/advisory designation such as CIA, US CPA, ACCA/FCCA or ICAEW ACA/FCA, before you can register
Comfort with general audit methodology (risk-based planning, sampling, evidence, reporting). AAIA builds on it rather than teaching it
Week 1 · 10 h
AI governance and risk for auditors (Domain 1)
Explain AI model types (supervised, unsupervised, generative, agentic) and the requirements and considerations an auditor must check for each
Evaluate an AI governance programme against NIST AI RMF, ISO/IEC 42001 and the EU AI Act, including the inventory, risk appetite, policies and accountable owners
Audit AI risk management (risk identification, rating, treatment, acceptance) and how privacy and data governance programmes cover AI
Map ethics principles and regulatory obligations to testable control objectives
Hands-on: Pick a public AI policy (for example a company's published responsible-AI standard) and write five control objectives with test steps for it. For each, note whether you would test design, operating effectiveness or both.
Audit data management for AI, including data sourcing, labelling quality, lineage, bias in training data and retention
Assess the AI development life cycle and change management (model versioning, retraining triggers, approval gates, rollback)
Evaluate how AI outputs, impacts and decisions are supervised, including human-in-the-loop design, drift monitoring and performance thresholds
Review AI testing techniques (validation sets, bias testing, robustness and adversarial testing, red teaming), AI-specific threats (prompt injection, data poisoning, model theft) and AI incident response
Hands-on: Write an audit work program for a bank's credit-decisioning model. Cover change management, monitoring thresholds, retraining approval, bias testing and incident escalation, with at least 15 test steps and the evidence you would request for each.
Plan and scope an AI audit, including risk-based selection of AI systems from the inventory and the audit criteria to use
Choose sampling methods for AI outputs and decisions, and explain the limits of sampling probabilistic systems
Collect and assess evidence such as model cards, evaluation logs, data lineage records and monitoring dashboards
Use data analytics to re-perform tests on model outputs, and write findings that rate risk and recommend remediation
Hands-on: Download a public dataset with a protected attribute (for example the UCI Adult income data), train or reuse a simple classifier, and re-perform a fairness test (selection rate by group, four-fifths rule). Write the result up as a one-page audit finding with condition, criteria, cause, effect and recommendation.