Last updated:
AAISM preparation guide
AAISM manager mindset: find the owner and the stage, then choose
Direct answer
On AAISM questions you are the AI security manager. You set AI security policy, design and run controls, and oversee AI risk, but the accountable owner of the AI use case decides whether to accept its risk. When two options both sound responsible, find the owner, the life-cycle stage and the step still missing. The stronger option closes that gap within your role. A hands-on fix can be right when the scenario gives it to you.
ISACA's AAISM outline lists 22 tasks. They ask you to establish, design, implement, monitor, oversee and advise on AI security; none asks you to evaluate or audit it. The AAISM candidate builds and runs AI security for the enterprise.
AAISM is open only to active CISM or CISSP holders, so the management judgment you already use carries over. What changes is the subject: AI use cases, training data, AI-specific threats, AI vendors and human oversight of what AI systems take in and produce.
Start AAISM baselineFind the missing step before you choose
| What the scenario shows | The tempting option | The AI security manager's step |
|---|---|---|
| Staff use an AI tool nobody approved | Block every AI tool at the proxy | Inventory the use, name an owner and bring it under the AI acceptable use policy |
| A model is ready for release, but nobody owns it after go-live | Run one more accuracy test | Name the owner, thresholds and escalation route for live monitoring before release |
| A vendor's AI service passed the vendor's own tests | Accept the vendor's report as proof | Ask for evidence that fits your own use, and contract terms for change notice |
| An AI risk is above the agreed threshold | The security team accepts it to keep the project moving | Give the accountable owner the treatment options; escalate if they would accept it outside the risk appetite |
| An attack through crafted prompts is under way | Wait for the next risk review | Contain it under the AI incident process, then notify and escalate as that process requires |
Original teaching example
An internal AI assistant answers staff questions from company documents. A tester shows that a crafted prompt makes it reveal HR files. The platform engineer offers to add a keyword filter tonight. As the AI security manager, what should you do first?
The filter treats one prompt, not the cause: the assistant can reach files its approved use never needed. First contain the exposure: remove the assistant's access to the HR files and tell its owner. Then fix the cause: classify the data sources, limit the assistant to the ones its use needs, and check whether anyone has already seen the files. The engineer's filter can be part of the fix, but it does not decide what the assistant may reach.
The AI security manager's job
Build and run AI security; leave the business decision with its owner.
ISACA's tasks include setting AI security policy, designing and reviewing AI security controls, overseeing the AI risk management life cycle, handling AI incidents and advising on AI architecture. Most AAISM scenarios give you an AI use, a fact that shows a gap, and a question about what the AI security manager should do.
Name the owner first
The accountable owner of the AI use case decides whether to accept its risk.
An option where the security team accepts an AI risk, or approves a business use on the owner's behalf, is usually the weaker answer. Your part is to make the risk visible against the agreed threshold, offer treatment options and escalate when a decision falls outside the risk appetite.
Buying the AI from a vendor does not move that decision. The organization that deploys it still owns the use and needs evidence that fits it.
Find the life-cycle stage
A sound control at the wrong stage is still the weaker answer.
Place the scenario in the AI life cycle: design, data, training, validation, deployment, monitoring or retirement. Then prefer the step that closes the gap at that stage, even when it sends work back to an earlier one. ISACA's subarea 3B covers the AI life cycle, including model selection, training and validation.
Check the oversight
A person in the loop counts only if they can actually override.
ISACA's tasks include risk-based human oversight of AI inputs and outputs. When an option adds a reviewer, check that the reviewer has the information, time and authority to override, and that someone owns the thresholds for drift and misuse after go-live.
When the technical answer is right
When the scenario gives the action to you, or the question turns on the control itself.
AI technologies and controls is 38% of the exam, and some questions turn on a specific control, such as input and output controls against prompt injection or data controls against poisoning. A live incident also needs action, not a committee: containment comes first.
CertArc's AAISM review names one kind of miss “Answered as a technician, not an AI security manager”. It is not a rule that technical answers are always wrong.
Three short scenarios
Each one turns on the owner, the stage or the oversight, not the technology.
- A product team wants to launch an AI feature next week, and the security review has open findings. Stronger: give the owner the open risks and the conditions for launch, and let them decide within the risk appetite, rather than signing it off yourself or blocking it without a decision.
- Staff paste customer data into a public AI chatbot. Stronger: bring the use under the AI acceptable use policy and awareness training, with an approved tool if the business needs one, rather than only adding a block rule.
- A fraud model's alerts drop sharply a month after release. Stronger: check the drop against the monitoring thresholds and send it to the model's owner for reassessment, rather than retraining the model at once.
How CertArc practice uses this
The review shows why the manager's option beats the tempting one.
After an AAISM question, CertArc compares your choice with the stronger option and shows why each tempting answer is weaker. The CertArc AAISM Lens walks the decision from scenario signal to winning move. Tag why you missed, so Pattern Intelligence can show which habit keeps catching you.
The AAISM lens in one sentence
Name who owns the AI use case, which life-cycle stage it is in and what oversight it needs, then choose the step within the AI security manager's role that closes the gap at that stage.
Sources and fact check
Source checked: 2026-10-22
- Study methods and examples are independent CertArc teaching guidance. ISACA's pages are the authority for exam rules and fees, which can change.
- Practice results do not convert to an official score or guarantee a pass. CertArc is not affiliated with or endorsed by ISACA.
Frequently asked questions
What does “think like a manager” mean on AAISM?
Choose at the AI security manager's level: policy, controls and oversight for an AI use, with the decision to accept its risk left to the accountable owner. It does not mean avoiding technical detail.
Is the AAISM mindset the same as CISM's?
The core carries over: risk ownership, business objectives and sequence. AAISM adds the facts that only exist with AI: the life-cycle stage, the training data, the vendor's part and human oversight of AI outputs.
How is it different from the AAIA auditor mindset?
AAIA asks what the auditor should evaluate and report. AAISM asks what the AI security manager should decide, build or run. The same scenario can have a different best answer in each.
Should I always choose the governance answer?
No. Choose the step the scenario's role, owner and stage support. AI technologies and controls is 38% of the exam, and some questions turn on a specific control.
Who accepts AI risk in an AAISM scenario?
In CertArc's teaching, the accountable owner of the AI use case, within the risk appetite. The AI security manager assesses the risk, proposes treatment and escalates.
Does a vendor's AI change who owns the decision?
No. The organization that deploys a vendor's AI still owns the decision to use it, and needs evidence that fits its own use.
