An aerospace manufacturing company that prioritizes accuracy and security has decided to use generative AI to enhance operations. Which of the following large language model (LLM) adoption plans BEST aligns with the company's risk appetite?
Answer(s): A
Developing a private LLM to automate non-critical functions best aligns with a high accuracy and security risk appetite by maintaining full control over data, model behavior, and security boundaries while limiting potential impact if issues arise.
During red team testing of an AI system used to make lending decisions, which of the following techniques BEST simulates a data poisoning attack?
Answer(s): D
Corrupting training data sets directly simulates a data poisoning attack by intentionally manipulating the data the model learns from, which can alter decision logic and bias outcomes such as lending approvals.
Which of the following is the MOST effective strategy for penetration testers assessing the security of an AI model against membership inference attacks?
Answer(s): B
Analyzing AI model confidence scores helps determine whether the model behaves differently for data it was trained on versus unseen data, which is the core signal exploited in membership inference attacks.
When creating a use case for an AI model that provides sensitive decisions affecting end users, which of the following is the GREATEST benefit of using model cards?
Model cards provide transparent documentation of ethical considerations, limitations, intended use, and potential impacts of the AI model, which is critical when the model makes sensitive decisions affecting end users.
The PRIMARY goal of data poisoning attacks is to:
Post your Comments and Discuss ISACA AAISM exam prep with other Community members:
Our website is free, but we have to fight against AI bots and content theft. We're sorry for the inconvenience caused by these security measures. You can access the rest of the AAISM content, but please register or login to continue.