Which of the following BEST describes the role of transparency in AI?
Answer(s): D
Transparency in AI is about making the system’s behavior and decision rationale understandable to relevant stakeholders, enabling clear explanations for how and why outcomes were produced.
During the deployment of a generative AI platform, a risk assessment highlighted threats such as data leakage and prompt manipulation. Which of the following is the BEST way to ensure appropriate control selection?
Answer(s): A
Mapping the identified threats to the organization’s existing control catalog ensures controls are selected based on the specific risks found, fit the enterprise governance model, and highlights where AI-specific safeguards are needed to close gaps (such as prompt injection and data leakage controls).
A school district contracts a third-party provider for AI-based curriculum recommendations. Which of the following is the BEST way to ensure the vendor uses AI responsibly?
Answer(s): C
A model card documents how the model was developed and evaluated, intended uses and limitations, key performance and bias considerations, and other responsible-AI details, giving the district the best basis to assess whether the vendor’s AI is being used and managed responsibly.
A military contractor discovered that its large language model (LLM) is at high risk of being targeted by advanced persistent threat (APT) actors seeking to exploit the model in order to access confidential information. Which of the following attacks is the HIGHEST priority to protect against?
Model inversion is the most urgent risk because it directly targets confidentiality by attempting to extract sensitive information from the model based on its outputs, potentially revealing confidential or proprietary data that the model learned during training or fine-tuning.
A model stealing attack has been launched on an AI system that has been in use for two years. Which of the following would BEST mitigate the impact of the attack?
Answer(s): B
Model stealing attacks rely heavily on access to model outputs, especially confidence scores, to reconstruct or approximate the model. Replacing the model with one that hides or limits confidence information directly reduces the attacker’s ability to extract useful signals, thereby mitigating the effectiveness and impact of the ongoing attack.
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.