AAISM Cheat Sheet: Key Facts to Review

The short version

This sheet is a quick review, not a full course. It lists the exam facts, the key ideas in each of the three domains, term pairs that people mix up, rules of thumb for scenario questions, and a checklist for the day before. It is written in our own words and does not claim to cover everything on the exam. Use it after you have studied, and use ISACA’s content outline as the list of what is in scope.

Exam facts at a glance

Item	Fact (ISACA, checked October 22, 2026)
Questions	90 multiple-choice, one best answer
Time	150 minutes
Passing score	450 on a scale of 200 to 800
Wrong answers	No penalty, so answer everything
Domains	31% governance, 31% risk, 38% technologies and controls
Prerequisite	Active CISM or CISSP
Retakes	Wait 30 days, then 90, then 90; four attempts in 12 months

More detail is on the exam format page.

Domain 1: AI governance and program management (31%)

Key ideas

• Governance sets direction and accountability. Management carries it out.
• AI needs its own policies, but they sit inside the existing policy hierarchy: policy, then standard, then procedure.
• An AI asset inventory covers models, datasets, prompts, tools, and AI features inside purchased software.
• Incident response for AI needs extra triggers such as harmful output or model drift.

Easily confused pairs

• Policy is the mandatory statement of intent. A standard defines required settings. A procedure gives steps.
• A framework is a reference structure. A regulation is binding law.
• Business continuity is about keeping the service going. Incident response is about handling the event.

Decision rules

• If a question asks what to do FIRST in a new AI initiative, think stakeholders, scope, and ownership before technical controls.
• If the scenario shows no owner for an AI system, the answer is usually to assign accountability.

Domain 2: AI risk management (31%)

Key ideas

• Risk is the combination of likelihood and impact. A threat exploits a vulnerability.
• Treatment options are accept, mitigate, transfer, and avoid.
• Vendor and supply chain risk covers suppliers of models, data, and hosted AI services.

Easily confused pairs

• Risk appetite is the broad level the organization will pursue. Risk tolerance is the acceptable variation around it.
• A risk assessment looks at risk to the organization. An impact assessment looks at effects on people or rights.
• A threat is the source. A vulnerability is the weakness. A risk is the possible loss.
• KPI measures performance. KRI signals rising risk. KCI measures whether controls work.

Decision rules

• If the supplier will not share information, think contractual rights, assurance, and a risk decision by the owner.
• If the risk exceeds tolerance, treat it or escalate it. Do not quietly accept it.

Domain 3: AI technologies and controls (38%)

Key ideas

• Training builds the model. Inference uses it. Different threats apply to each.
• Data controls cover quality, lineage, labeling, classification, and retention.
• Monitoring watches for drift, misuse, and unusual output.
• Human oversight should match the risk of the decision.

Easily confused pairs

• Data poisoning corrupts training data. Prompt injection manipulates input at use time.
• Model inversion tries to recover training data. Membership inference tests whether a record was in the training set.
• Data drift means inputs change. Concept drift means the relationship between inputs and outcomes changes.
• A datasheet describes a dataset. A model card describes a model.
• Input filtering screens what goes in. Output filtering screens what comes out.

Decision rules

• If a retrieval system returns documents, enforce the user’s access rights before the model sees them.
• If an agent has more access than its task needs, reduce its permissions.
• If a control exists but nobody checks it, add monitoring.

Common traps

• Picking a technical fix when the question asks for a management action.
• Ignoring the qualifier (BEST, FIRST, MOST likely) and choosing a true but less suitable answer.
• Treating a framework name as a requirement. Frameworks guide, laws bind.
• Choosing “stop using AI” when a proportionate control would do.
• Mixing up assessing a risk with treating it.

Reading a scenario question

1. Underline the qualifier.
2. Note the role you are playing (manager, architect, advisor).
3. Find the constraint, such as budget, law, or time.
4. Remove answers that are true but do not fit the constraint.
5. Pick the one that a responsible manager would do next.

Last-day checklist

• Check your ID, your name match, and your appointment time and language.
• For remote exams, test the room, camera, and connection.
• Review this sheet once, then stop.
• Plan your pacing: about 100 seconds per question, with a check at the halfway mark.
• Remember the single ten-minute break and that the timer keeps running.
• Sleep. A rested mind helps more than one more page.

More guides for AAISM

• AAISM Study Plan: 4-Week and 8-Week
• Is the AAISM Exam Hard?
• AAISM Prerequisites: CISM or CISSP First
• AAISM Renewal and Retakes Explained
• Is AAISM Worth It? An Honest Look
• AAIA vs AAISM: Which ISACA AI Exam First?
• AAIR vs AAISM: Which ISACA AI Exam First?

Next step

Test your recall with the sample questions and go deeper in domains explained. When you want a full check, our AAISM practice tests offer five sets of 90 original questions. Return to the overview any time.