A regulator asks an insurance company to prove that a specific decision made by its underwriting model was not based on a prohibited characteristic (e.g., race). The company has the model's code and the final output, but cannot show how the input data was transformed or which features were used in the final decision. What is the MOST critical missing capability?

A more accurate model with a higher prediction score.

A faster data pipeline to process applications more quickly.

A backup of the production database containing the raw application data.

End-to-end data lineage that documents every transformation from raw data to model input.

A multinational corporation has just completed a merger. The combined entity has multiple AI assets from both companies, including models, datasets, and supporting infrastructure. The CISO is concerned about shadow AI and wants to establish a unified governance program. The legal department has identified that some of the acquired company's AI assets process data subject to GDPR, and some process data subject to CCPA. The CISO wants to prioritize the governance effort. What is the MOST appropriate first step?

Focus only on the AI assets that process GDPR data, as GDPR has the most stringent requirements.

Migrate all AI assets to a single cloud platform to simplify management.

Conduct a comprehensive discovery and inventory of all AI assets from both companies, then classify them by criticality, sensitivity, and regulatory impact.

Immediately shut down all AI assets from the acquired company until they can be fully reviewed.

A fraud detection model has been in production for two years. The model's performance has been declining over the past three months, and the data science team suspects that the characteristics of fraudulent transactions have changed. What is the MOST appropriate action to address this situation?

Increase the frequency of model retraining to weekly to keep the model current.

Roll back to the previous version of the model that had better performance.

Retrain the model immediately with the most recent data and deploy it to production.

Investigate the model's performance metrics and compare the current input data distribution to the training data distribution to confirm data drift.

Which of the following is a reason to monitor AI assets in production?

To detect performance degradation and security issues

To change the model's programming language

To increase the number of training epochs

To reduce the size of the model

ISACAAdvanced in AI Security Management

Domain 1Objective 3

AI Asset and Data Life Cycle Management AAISM Practice Questions (Page 3)

Part of the AI Governance and Program Management domain, which accounts for 31% of the AAISM exam.

45questions here

9free pages

15concepts

31%of the exam

15 Concepts in This Objective

AI Asset Identification
AI Asset Classification
AI Asset Inventory Management
AI Asset Life Cycle Stages
Data Life Cycle Management
Data Governance for AI
Data Lineage and Provenance
Data Privacy and Protection
Data Retention and Disposal
AI Model Life Cycle Management
Model Versioning and Reproducibility
AI Asset Risk Assessment
AI Asset Compliance and Audit
AI Asset Monitoring and Maintenance
AI Asset Retirement and Decommissioning

Questions 11–15

11foundation · easy
During which stage of the AI asset life cycle is the model typically monitored for performance degradation?
Select an answer first
12application · medium
A regulator asks an insurance company to prove that a specific decision made by its underwriting model was not based on a prohibited characteristic (e.g., race). The company has the model's code and the final output, but cannot show how the input data was transformed or which features were used in the final decision. What is the MOST critical missing capability?
Select an answer first
13expert · hard
A multinational corporation has just completed a merger. The combined entity has multiple AI assets from both companies, including models, datasets, and supporting infrastructure. The CISO is concerned about shadow AI and wants to establish a unified governance program. The legal department has identified that some of the acquired company's AI assets process data subject to GDPR, and some process data subject to CCPA. The CISO wants to prioritize the governance effort. What is the MOST appropriate first step?
Select an answer first
14application · medium
A fraud detection model has been in production for two years. The model's performance has been declining over the past three months, and the data science team suspects that the characteristics of fraudulent transactions have changed. What is the MOST appropriate action to address this situation?
Select an answer first
15foundation · easy
Which of the following is a reason to monitor AI assets in production?
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 Asset and Data Life Cycle Management AAISM Practice Questions (Page 3)

Questions 11–15

During which stage of the AI asset life cycle is the model typically monitored for performance degradation?

A fraud detection model has been in production for two years. The model's performance has been declining over the past three months, and the data science team suspects that the characteristics of fraudulent transactions have changed. What is the MOST appropriate action to address this situation?

Which of the following is a reason to monitor AI assets in production?

Keep studying — AI Governance and Program Management