Last updated:
AAISM preparation guide
AAISM Domain 1: AI governance and program management
Direct answer
Domain 1 asks whether you can run AI security as a business program: who decides, which policy applies, what is inventoried, and how incidents involving AI are handled. The best answer usually sets ownership and governance before a technical fix.
Experienced security managers often know the control but miss the governance step: an AI use case without an owner, a generative AI tool without an acceptable-use policy, or a model nobody has inventoried.
Read each scenario for the decision-maker and the stage the program is at. A sound control can still be the wrong answer if the use case has not been approved or scoped.
Start AAISM baselineDomain 1 subareas (31% of the exam)
| Subarea | What the questions test |
|---|---|
| 1A Stakeholder considerations, industry frameworks and regulatory requirements | Who must be involved, which framework or regulation applies, and how roles are assigned |
| 1B AI-related strategies, policies and procedures | Business cases, AI strategy, acceptable use, policy development and ethics |
| 1C AI asset and data life cycle management | Inventorying AI assets and the data they use, from acquisition to retirement |
| 1D AI security program development and management | Building and running the program, its components, and AI-enabled security |
| 1E Business continuity and incident response | Preparing for, identifying, assessing, responding to and reviewing AI incidents |
Original teaching example
A team wants to launch a customer-facing chatbot next month. It will answer customers with text generated by a vendor's model, drawing on customer records, so it could state terms the company never offered or repeat one customer's details to another. No policy covers generative AI, and legal has not reviewed the data the model may see. What should the AI security manager do first?
First, escalate to the accountable business owner, with legal and privacy, and recommend holding the launch until a generative AI policy, a risk assessment of the use case and a review of the vendor's terms on data use are in place. The design should then limit the model to the authenticated customer's own records and ground its answers in approved terms. A technical guardrail alone treats a policy gap as a configuration task.
Incidents involving AI still follow the response sequence
Prepare, identify and report, assess, respond, then review. What changes is what you contain and preserve: the model or feature, prompts, logs and model versions.
Use CISM judgment, then check the AI-specific fact
Business alignment and risk ownership still decide many questions. Look for the fact that only exists because AI is involved, such as an unapproved use case or a missing human override.
Frameworks and regulation
The outline names industry frameworks and regulatory requirements without listing specific ones.
Subarea 1A tests applying the frameworks and regulations that bind an organization, not reciting a particular one. CertArc does not claim that any named framework is tested.
Worked examples by subarea
One original scenario per subarea, with the stronger answer.
- 1A: A product team plans an AI feature that scores customers' credit applications and has not checked AI-specific regulation. Stronger: identify the AI-specific duties that apply, for example whether the use is classed as high risk, and the stakeholders who must approve, before the design is fixed.
- 1B: Staff paste client data into a public generative AI chatbot whose terms allow prompts to be used for training. Stronger: a generative AI acceptable-use policy with approved tools and rules for prompt data, not only a block on one site.
- 1C: A model is retrained on a new customer dataset, and nobody can say which model version is in production or which data trained it. Stronger: an AI asset inventory that links each model version to its training data, owner and approved use, so a data problem can be traced to the models it affected.
- 1D: Business units adopt generative AI tools and fine-tune models on customer data on their own, with no common rules for training data, model evaluation or AI incident reporting. Tempting: a blanket block on AI tools. Stronger: an AI security program with model approval based on evaluation evidence, rules for training data, AI incident reporting, owners and metrics.
- 1E: After a model update, a chatbot starts revealing other customers' details, and the team rolls back and deletes the prompt logs. Stronger: contain the feature and preserve prompts, outputs and the model versions before and after the update, so the cause can be found.
Sources and fact check
Source checked: 2026-10-12
- 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
Is Domain 1 just CISM governance again?
It uses the same management judgment, applied to AI use cases, AI policy and AI incidents.
Do I need to memorise AI regulations?
Know what common frameworks and regulations require of an organization; questions test applying them, not reciting them.
How is Domain 1 weighted?
31% of the exam, by ISACA's outline.
How many subareas does Domain 1 have?
Five, 1A to 1E, in ISACA's outline.
Where should a CISM holder focus in Domain 1?
On the AI-specific facts: AI use-case ownership, generative AI policy, the AI asset inventory and AI incidents.
