ISACA Advanced in AI Security Management (AAISM) - AAISM Free Exam Questions

A. The cost of noncompliance was not determined

B. The AI tool is widely used within the industry

C. The AI tool is regularly audited

D. The risk is within the organization's risk appetite

A. Verifying the vendor has updated terms of service

B. Confirming the AI solution supports single sign-on (SSO)

C. Requiring the vendor to provide the model card

D. Ensuring the vendor offers 24/7 technical support

A. Conduct an AI impact assessment, evaluating acceptable levels of autonomy.

B. Expand insurance coverage to account for possible patient lawsuits.

C. Create awareness to highlight the AI use cases and benefits.

D. Align AI infrastructure to existing healthcare enterprise architecture (EA).

A. Use the latest version of all libraries from public repositories

B. Maintain a list of frequently used libraries to ensure consistent application in projects

C. Retrain the model regularly to handle package and library updates

D. Scan the packages and libraries for malware prior to installation

A. Response time and throughput

B. Error rate and bias detection

C. Explainability and F1 score

D. Customer effort score and user retention rate

A. Purchasing an LLM dataset on the open market

B. Developing a public LLM to automate critical functions

C. Contracting LLM access from a reputable third-party provider

D. Developing a private LLM to automate non-critical functions

A. Excessive reliance on external consultants for model design

B. Absence of metrics and dashboards for analysts

C. Lack of sufficient computing resources for the AI system

D. Insufficient model validation and change control processes

A. Industry benchmarking peer review

B. Third-party audit reports

C. Internal red team testing reports

D. General AI security whitepapers

A. Limiting API quotas per user

B. Rolling back to a previous model checkpoint

C. Enforcing validation before persistence

D. Disabling reflection patterns

A. Detecting irregular pixel patterns or distorted frames

B. Analyzing biometric inconsistencies

C. Comparing the model card to the output generated by AI

D. Examining irregularities in metadata or data traces

A. Require customer service to manually price and approve orders until the AI system is fully restored.

B. Revert to the previous AI model and validate backup of the feature store.

C. Failover to a secondary environment and retrain the model with recent data.

D. Keep the system offline and conduct a full forensic investigation.

A. Implementing a kill switch

B. Validating AI model training data

C. Monitoring AI outputs against policy

D. Ensuring controls exceed industry benchmarks

A. Prompt sanitization

B. Model fine tuning with increased use of function calls

C. Allow list employee access

D. Retrieval-augmented generation (RAG) with access control lists (ACLs)

A. Implementing continuous monitoring of the system

B. Requiring change control notifications in the contract

C. Including a human-in-the-loop

D. Conducting penetration testing