# ISACA AAISM: study planner

Domain weights, the exam length and the exam rules come from ISACA's AAISM Exam Candidate Guide, version 1.26, with the Examination Content Outline of June 2025, checked against ISACA's published copy on 14 November 2026. The certification and CPE rules come from ISACA's Get AAISM Certified page, read on 16 November 2026. ISACA does not recommend a number of study hours: the budgets and the split are our suggestion, in proportion to the published weights. Check the current outline before you start in case the weights have changed.

## Hours by domain

| Domain | Share | 50 hours | 100 hours | 150 hours |
|---|---|---|---|---|
| 1. AI Governance and Program Management | 31% | 15.5 | 31 | 46.5 |
| 2. AI Risk and Opportunity Management | 31% | 15.5 | 31 | 46.5 |
| 3. AI Technologies and Controls | 38% | 19 | 38 | 57 |
| **Total** | **100%** | **50** | **100** | **150** |

ISACA publishes a share per domain and nothing finer. It does not weight the subdomains, so the rows below have no hours of their own: spread each domain's hours across its subdomains by how new the material is to you.

## Pace for timed practice

| Set | Questions | Time at the exam's average pace |
|---|---|---|
| One question | 1 | 1 minute 40 seconds |
| Short timed set | 10 | 16 minutes 40 seconds |
| Half a paper | 45 | 1 hour 15 minutes |
| Full paper | 90 | 2 hours 30 minutes |

The exam is 90 questions in 150 minutes, which is 100 seconds a question on average. Some stems are a paragraph of scenario, so bank time on the short ones.

## Practice questions (our suggestion)

Our AAISM Practice Questions carries 200 questions, sized to the published shares: 62 for Domain 1, 62 for Domain 2 and 76 for Domain 3. Those counts are our arithmetic on ISACA's percentages, not a published item count.

- [ ] Each domain's questions worked untimed, straight after reading that domain in the guide
- [ ] Each domain's questions sat again at 100 seconds a question
- [ ] Every rationale read, including the ones for questions you got right
- [ ] In the final week, all 200 sat in mixed order across a few sessions
- [ ] Every miss logged against its subdomain, and the guide reread there

## 1. AI Governance and Program Management

31% of the exam.

| Subdomain | Read | Studied | Practiced |
|---|---|---|---|
| 1A Stakeholder considerations, industry frameworks and regulatory requirements | [ ] | [ ] | [ ] |
| 1B AI-related strategies, policies and procedures | [ ] | [ ] | [ ] |
| 1C AI asset and data life cycle management | [ ] | [ ] | [ ] |
| 1D AI security program development and management | [ ] | [ ] | [ ] |
| 1E Business continuity and incident response | [ ] | [ ] | [ ] |

- [ ] Hours in the calendar: 15.5, 31 or 46.5
- [ ] Can name every stakeholder for an AI use case, and spot the one missing from a decision (legal, privacy, the business owner)
- [ ] Can say what the AI governance charter settles: who approves a use case, who accepts its risk, who can stop a deployment
- [ ] Can tell a policy from a standard, a procedure and a guideline, and say which layer to fix when staff break a clear rule
- [ ] Can build an AI inventory that counts use cases as well as models, including AI switched on inside licensed software
- [ ] Can follow a deletion to every copy: training snapshots, embeddings, vector indexes, backups
- [ ] Can measure whether awareness training works with evidence from controls, not completion rates
- [ ] Can put AI dependencies, silent failure and provider model retirement into a business impact analysis
- [ ] Can order an AI incident: contain first, preserve evidence, then eradicate, recover to a known-good version, and learn

## 2. AI Risk and Opportunity Management

31% of the exam.

| Subdomain | Read | Studied | Practiced |
|---|---|---|---|
| 2A AI risk assessment, thresholds and treatment | [ ] | [ ] | [ ] |
| 2B AI threat and vulnerability management | [ ] | [ ] | [ ] |
| 2C AI vendor and supply chain management | [ ] | [ ] | [ ] |

- [ ] Hours in the calendar: 15.5, 31 or 46.5
- [ ] Can scope a risk assessment to the use case, not the model
- [ ] Can turn a model metric into a risk question: volume, direction of errors, results by segment, controls
- [ ] Can write a threshold a person can measure, and say who sets it and who accepts what sits above it
- [ ] Can tell mitigation, transfer, avoidance and acceptance apart, and say when another control is over-treatment
- [ ] Can name the main attacks on AI: poisoning, evasion, extraction, inversion, membership inference, prompt injection, jailbreaks
- [ ] Can explain why one clean red team run proves little, and what an attack success rate is
- [ ] Can say what a vendor's certificate or assurance report covers, and when a change means reopening the assessment
- [ ] Can find the fourth party behind an embedded AI feature, and the data-use terms that matter

## 3. AI Technologies and Controls

38% of the exam, the largest domain.

| Subdomain | Read | Studied | Practiced |
|---|---|---|---|
| 3A AI security architecture and design | [ ] | [ ] | [ ] |
| 3B AI life cycle (model selection, training and validation) | [ ] | [ ] | [ ] |
| 3C Data management controls | [ ] | [ ] | [ ] |
| 3D Privacy, ethical, trust and safety controls | [ ] | [ ] | [ ] |
| 3E Security controls and monitoring | [ ] | [ ] | [ ] |

- [ ] Hours in the calendar: 19, 38 or 57
- [ ] Can draw the trust boundaries in an AI system and say why the model can neither decide nor enforce access
- [ ] Can apply least agency to an agent: narrow tools, scoped identity, approval for irreversible actions, limits in code, a kill switch
- [ ] Can compare building, fine-tuning, open-weight and hosted models from the security seat
- [ ] Can say what makes validation independent, and what changes after release make it stale
- [ ] Can protect weights and training checkpoints as copies of the model
- [ ] Can say why masked, tokenized or synthetic data is reduced risk, not no risk
- [ ] Can place human oversight in, on or over the loop, and test whether the human can change the outcome
- [ ] Can say what to log for models and agents, and how to tell a working detection from a quiet one
- [ ] At least half the hours spent on practice questions (our suggestion)

## Booking and exam day

- [ ] Active CISM, or active CISSP from ISC2, in hand (ISACA rule)
- [ ] Registered and paid: US$459 for members, US$599 for nonmembers; six months to sit from registration (ISACA rule)
- [ ] Appointment booked: from 48 hours after payment, and no more than 90 days ahead (ISACA rule)
- [ ] Any reschedule done at least 48 hours before the appointment (ISACA rule)
- [ ] Need longer? One six-month extension for US$75 (ISACA rule)
- [ ] ID ready: current, valid, original and government-issued, with your name, signature and photograph on one document; digital and military IDs are not accepted (ISACA rule)
- [ ] At a test center: arrive at least 30 minutes before the start; more than 15 minutes late and the fee is forfeit (ISACA rule)
- [ ] At home: device compatibility check done; room scan and mirror check ready; no breaks without the proctor's permission (ISACA rule)
- [ ] Closed book: no notes, no calculator, nothing on the desk (ISACA rule)
- [ ] Every question answered: there is no penalty for a wrong answer (ISACA rule)

## After the exam

- [ ] Preliminary result read on screen; official score expected within 10 working days (ISACA rule)
- [ ] If you passed: apply within five years, with the US$50 application processing fee (ISACA rule)
- [ ] If you did not: four attempts within a rolling 12 months; wait 30 days before the second attempt and 90 days before each one after (ISACA rule)
- [ ] Once certified: 10 hours of AI CPE each year from the next calendar year, your CISM or CISSP kept active, and ISACA's Code of Professional Ethics followed (ISACA rule)
