AAIR Exam Domains Explained With Weightings

The short version

The AAIR exam has three domains: AI Risk Governance and Framework Integration (37%), AI Life Cycle Risk Management (21%), and AI Risk Program Management (42%). The weightings and sub-area names come from the content outline in ISACA’s AAIR exam candidate guide, read on October 18, 2026. The explanations of what each area means in practice are ours. The traps are general observations about this style of exam, not statements from ISACA.

The percentages show how much of the exam comes from each domain. Domain results are for your information, and the passing score is based on the total.

Domain 1: AI Risk Governance and Framework Integration (37%)

This domain asks how an organization sets itself up to take AI risk on purpose. It has six sub-areas.

  • A. AI models, frameworks, strategies, and use cases. Know the main kinds of AI systems, the AI risk frameworks and standards in use, and how an AI strategy links to business goals. In practice: deciding whether a proposed use case fits the organization’s appetite.
  • B. Organizational processes and alignment. Fitting AI risk into existing processes, such as change management, the enterprise risk register, and architecture review.
  • C. Ownership, oversight, and accountability. Who is the risk owner, who challenges, who gives assurance. Think of the three lines model.
  • D. Policies, procedures, and training. What an AI policy should say, how exceptions work, and how to build AI awareness into training.
  • E. Regulatory compliance and legal considerations. Understanding how AI laws and standards classify systems and what duties follow, for example for providers versus deployers.
  • F. Trustworthiness, ethics, and societal impact. Bias, privacy, safety, transparency, and environmental, social, and governance (ESG) effects.

Typical traps. Mixing up appetite, tolerance, and capacity. Giving a risk to the team that built the model rather than the business owner. Assuming a contract moves accountability to a supplier. Treating a legal ban as something a policy exception can approve.

What to practice. Write one-sentence definitions of the core terms. Practice classifying use cases by risk level. Practice deciding who owns what in a scenario.

Domain 2: AI Life Cycle Risk Management (21%)

This is the smallest domain, and it follows an AI system from idea to retirement. It has four sub-areas.

  • A. Design, development or procurement, and documentation. Risks in choosing to build or buy, in requirements, and in documentation such as model descriptions and limits of use.
  • B. Model training, testing, and validation. Data quality, bias, overfitting, evaluation by subgroup, and independent validation.
  • C. Implementation, maintenance, and decommissioning. Monitoring for drift, change control, retraining, and retiring a model safely.
  • D. Data and asset management. Inventories of models and data, data lineage, access, and retention.

Typical traps. Picking a technical fix when the question wants a governance step, or the reverse. Confusing similar problems: data drift, concept drift, overfitting, and data leakage are different things. Trusting one overall accuracy number when results differ by group.

What to practice. For each life cycle stage, name one risk, one control, and one piece of evidence that shows the control works. Learn the difference between the terms listed above well enough to spot them in a short scenario.

Domain 3: AI Risk Program Management (42%)

The biggest domain covers the daily work of a risk program. It has six sub-areas.

  • A. Risk scenario identification and assessment. Writing usable scenarios with a threat, a weakness, and an impact, including attacks specific to AI such as data poisoning or prompt injection, and choosing a qualitative or quantitative approach.
  • B. Risk treatment strategies. Mitigate, transfer, avoid, or accept, and when each is defensible.
  • C. Controls management. Choosing preventive, detective, or corrective controls, then testing that they work.
  • D. Risk metrics, monitoring, and reporting. Risk indicators versus control indicators, thresholds, and tailoring reports to the board, management, or operations.
  • E. AI supply chain risk management. Vendor due diligence, contract terms, model version changes, and third-party data.
  • F. Incident response, business impact analysis, continuity, and disaster recovery. Handling an AI incident and planning for outages of AI-dependent processes.

Typical traps. Calling a measure of customer satisfaction a control indicator. Ranking risks by inherent score when controls already work. Choosing detection when prevention is possible and the harm cannot be undone. Choosing a perfect quantitative model when there is no data to support it.

What to practice. Write three scenarios of your own and take each through treatment, control, and metric. Practice questions that give two defensible answers and one constraint, and learn to let the constraint decide.

The skills across domains

ISACA’s outline also lists 23 skills, such as integrating AI risk into the enterprise risk framework, advising on AI clauses in contracts, assessing human oversight controls at critical decision points, and using AI to support the risk program itself. They cut across the three domains. Read the whole list once, because it shows the verbs the exam uses: evaluate, integrate, advise, assess, and monitor.

How our practice sets map to the domains

Each of our five practice sets has 90 questions split 33, 19, and 38 across the three domains, in line with the published weightings. Every question is tagged with its domain, so you can track which area costs you the most marks.

More guides for AAIR

Next step

Try the question style on the sample questions page, or go back to the study plan to schedule each domain.

Try these practice exams