AAISM Domains Explained

The short version

The AAISM exam has three domains: AI Governance and Program Management (31%), AI Risk Management (31%), and AI Technologies and Controls (38%). The sub-areas below come from ISACA’s exam content outline (the candidate guide dates it June 2025). The notes on traps and practice are our own general explanation, not ISACA statements.

Domain	Weighting	Sub-areas listed by ISACA
1. AI Governance and Program Management	31%	5
2. AI Risk Management	31%	3
3. AI Technologies and Controls	38%	5

ISACA also lists 22 task statements, such as defining the charter and roles for AI governance, conducting AI impact assessments, and providing risk-based human oversight of AI inputs and outputs. They describe the job, and they cut across the domains.

Domain 1: AI Governance and Program Management (31%)

What it covers

ISACA lists five areas:

Stakeholder considerations, industry frameworks, and regulatory requirements
AI-related strategies, policies, and procedures
AI asset and data life cycle management
AI security program development and management
Business continuity and incident response

What it means in practice

This is the “run it as a program” domain. You decide who is accountable for AI, what the policies say, how AI assets are found and recorded, how metrics reach management, and what happens when an AI system fails or is attacked. Regulation sits here too: you should know how to work out which rules apply to a system, for example by asking who built it, who deploys it, and what it is used for. Laws such as the EU AI Act and frameworks such as the NIST AI Risk Management Framework are common study material, though check the outline for what ISACA actually names.

Typical traps

Choosing a technical fix when the question is about ownership, policy, or sequence.
Mixing up governance (setting direction and accountability) with management (carrying it out).
Forgetting that an AI feature can arrive silently inside software you already own. The first step is usually to find it and record it.
Treating an AI incident as only an attack. Harm can occur when nobody attacks anything, for example through drift or biased output.

What to practice

Distinguish policy, standard, and procedure. Practice incident steps in order: contain, notify, escalate, eradicate, recover. Be able to say what belongs on an AI asset inventory.

Domain 2: AI Risk Management (31%)

What it covers

ISACA lists three areas:

AI risk assessment, thresholds, and treatment
AI threat and vulnerability management
AI vendor and supply chain management

What it means in practice

Here you rate and treat risk that is specific to AI. Risk shifts as data, usage, and the outside world change, so continuous monitoring matters more than a once-a-year review. You also deal with suppliers: hosted models, third-party datasets, and pre-trained models downloaded from public hubs. A supplier that will not share logs or training details is a risk you have to manage through contracts and assurance, not by hope.

Typical traps

Confusing a threat (an actor or event), a vulnerability (a weakness), and a risk (likelihood combined with impact).
Mixing appetite (how much risk leadership will take, set first) with tolerance (acceptable variation, expressed as thresholds).
Treating insurance or a contract clause as if it removes the risk. Transfer moves financial impact, not the likelihood of the event.
Mixing up a risk assessment with an impact assessment. They answer different questions.

What to practice

Rank treatment options for a given scenario. Identify internal versus external factors that trigger reassessment. List what a usable risk register entry needs: an owner, a treatment, and a status.

Domain 3: AI Technologies and Controls (38%)

What it covers

ISACA lists five areas:

AI security architecture and design
AI life cycle (for example model selection, training, and validation)
Data management controls
Privacy, ethical, trust, and safety controls
Security controls and monitoring

What it means in practice

This domain asks whether you can judge a design. Examples of the kind of topic you should be ready for: separating training from production, restricting an AI agent’s permissions, enforcing access control when data is retrieved rather than after the answer is written, checking the integrity of training data and labels, and monitoring a deployed model for drift. Privacy and safety controls belong here too, including human oversight and explainability.

Typical traps

Picking encryption for every question. Encryption protects confidentiality, but the problem may be integrity, access, or separation of duties.
Choosing detection when the question asks for prevention.
Filtering a model’s output when the real fix is to stop restricted data from reaching the model in the first place.
Forgetting trade-offs, such as robustness gained at the cost of accuracy on ordinary inputs.

What to practice

Match attack types to controls (poisoning, evasion, extraction, prompt injection). Separate neighboring artifacts such as a model card, a dataset datasheet, and a bill of materials. Practice reading a scenario for the one constraint that decides between two good options.

How to spend your time

Domain 3 has the largest weighting, but the two management domains add up to 62% of the exam. Engineers may be tempted to under-prepare Domains 1 and 2, and governance professionals Domain 3. Check your baseline results before you decide.

More guides for AAISM

Next step

Put the domains into a schedule with the study plan, or try the sample questions. Our AAISM practice tests cover all three domains with explanations on every answer option. For the current outline, see ISACA’s exam content outline.

Try these practice exams