A company is setting up an AI development environment that uses sensitive customer data. The environment must be isolated from the production network to prevent data leakage, but developers need to access the environment for testing. Which architecture best meets these requirements?

Use a separate virtual network with a VPN for developer access.

Use the production network with strict firewall rules.

Use a public cloud environment with no network restrictions.

Use a dedicated physical server with no network connectivity.

A startup is developing an AI-powered document summarization tool. The tool will be used by multiple clients, and each client's data must be kept separate. The architecture uses a shared model but individual client-specific fine-tuning. Which security principle is most critical to implement?

Defense-in-depth: implement multiple layers of security controls.

Secure-by-design: build security into the development process.

Least privilege: ensure each client can only access their own data and model.

Data minimization: collect only the data necessary for the service.

ISACAAdvanced in AI Security Management

Domain 3Objective 1

AI Security Architecture and Design AAISM Practice Questions (Page 6)

Part of the AI Technologies and Controls domain, which accounts for 38% of the AAISM exam.

31questions here

7free pages

7concepts

38%of the exam

Questions 26–30

26foundation · medium
Which type of attack on an AI system involves an attacker deliberately introducing malicious data into the training dataset to corrupt the model's behavior?
Select an answer first
27application · medium
A company is setting up an AI development environment that uses sensitive customer data. The environment must be isolated from the production network to prevent data leakage, but developers need to access the environment for testing. Which architecture best meets these requirements?
Select an answer first
28foundation · medium
Which AI-specific threat involves an attacker querying a model to reconstruct or infer sensitive information from the training data?
Select an answer first
29application · medium
A startup is developing an AI-powered document summarization tool. The tool will be used by multiple clients, and each client's data must be kept separate. The architecture uses a shared model but individual client-specific fine-tuning. Which security principle is most critical to implement?
Select an answer first
30foundation · medium
Which principle of secure AI architecture ensures that even if one security control fails, additional layers of protection remain to prevent or mitigate an attack?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to AAISM

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by ISACA. “AAISM” is a trademark of its owner, used for identification only.

AI Security Architecture and Design AAISM Practice Questions (Page 6)

Questions 26–30

Which type of attack on an AI system involves an attacker deliberately introducing malicious data into the training dataset to corrupt the model's behavior?

A company is setting up an AI development environment that uses sensitive customer data. The environment must be isolated from the production network to prevent data leakage, but developers need to access the environment for testing. Which architecture best meets these requirements?

Which AI-specific threat involves an attacker querying a model to reconstruct or infer sensitive information from the training data?

A startup is developing an AI-powered document summarization tool. The tool will be used by multiple clients, and each client's data must be kept separate. The architecture uses a shared model but individual client-specific fine-tuning. Which security principle is most critical to implement?

Which principle of secure AI architecture ensures that even if one security control fails, additional layers of protection remain to prevent or mitigate an attack?

Keep studying — AI Technologies and Controls