AMAADOR ACADEMYCertificatesVerify a certificate

AAISM

Final check — ISACA AAISM Exam Prep: Advanced in AI Security Management Course and Mock Exam

Answer the 42 questions. You need 80% or more to pass. You can retry as many times as you like.

Study the course →

Question 1Which credential must a candidate hold to be eligible for the ISACA AAISM exam?
Question 2Which AAISM domain carries the highest weight in the current exam content outline?
Question 3A board member asks the CISO which AI framework the company could be certified against to show customers it manages AI responsibly. What should the CISO name?
Question 4A retailer learns that a new AI law will apply to some of its systems next year. It has an AI inventory with owners. What should the security manager do NEXT?
Question 5A company builds its own customer chatbot and offers it to the public in the EU. Which EU AI Act obligation applies MOST directly to this use?
Question 6In the NIST AI RMF, which function is cross-cutting and informs the other three?
Question 7Business units keep launching AI pilots that security learns about only after go-live. What is the BEST long-term fix?
Question 8Which item is MOST important to record for each entry in an AI asset inventory?
Question 9A bank decides to retire a credit model that was trained on customer data. What should the retirement procedure include?
Question 10A CISO is building an AI security program. The company already has an ISO/IEC 27001 ISMS. What is the BEST approach?
Question 11A hospital’s triage assistant uses a third-party AI API. The business impact analysis rates triage as critical. What should the continuity plan include?
Question 12After a poisoning attack is confirmed on a fraud model, the team has contained the incident. What should happen before the model returns to production?
Question 13Two AI use cases use the same model: one drafts marketing copy for human editing, the other approves small refunds automatically. Which factor MOST explains the difference in risk?
Question 14Who should formally accept the residual risk of a new AI use case?
Question 15A team proposes to set fairness thresholds for a lending model after reviewing the first test results. What is the main problem?
Question 16A proposed AI system would score citizens’ social behaviour to decide access to public services in the EU. What is the BEST risk treatment?
Question 17In the NIST AI RMF, which function covers understanding the context, purpose and potential impacts of an AI system?
Question 18An insurer signs a contract in which the AI vendor indemnifies it for losses caused by model errors. Which statement is correct?
Question 19An LLM summarises web pages for analysts. A page contains hidden text telling the model to insert a phishing link in its summary. What attack is this?
Question 20Monitoring shows one API customer sending millions of systematically varied queries to a proprietary pricing model. What is the MOST likely threat?
Question 21Which item is ranked first in the 2025 OWASP Top 10 for LLM Applications?
Question 22Developers want to download an open-source model from a public hub and load it in production. What should the security manager require?
Question 23During due diligence on a generative AI SaaS tool, which question is MOST important for confidentiality?
Question 24What is the main purpose of an AI bill of materials (AI-BOM)?
Question 25An LLM generates SQL queries from user questions, and the application runs them directly on the customer database. What is the BEST control?
Question 26Why should the system prompt NOT be treated as a security boundary?
Question 27A data science team trains a model and also runs the only validation tests before approving release. What is the main weakness?
Question 28A company chooses between fine-tuning an open model and buying an API-based model. What should drive the decision FIRST?
Question 29Which practice BEST supports rollback after a faulty model release?
Question 30An AI agent can browse the web, read files and send payments. In threat modelling, which design choice MOST reduces impact if it is manipulated?
Question 31An investigator suspects a training dataset was tampered with months ago. Which control would help MOST to confirm it?
Question 32How should a vector database built from confidential documents be classified?
Question 33A lender’s model automatically rejects loan applications with no human involvement. Customers in the EU complain. Which control BEST addresses their rights?
Question 34Bias testing shows a recruitment model scores women lower even though gender is not an input. What is the MOST likely cause?
Question 35Which control BEST turns an AI ethics principle into something auditable?
Question 36A media company in the EU publishes realistic AI-generated images showing real public figures (deep fakes). Which transparency control is MOST relevant?
Question 37A developer edits the system prompt of a live chatbot directly in production to fix a tone issue. What control was bypassed?
Question 38A fraud model’s inputs start to differ from training data after a new payment channel launches, but no alert fires. What is missing?
Question 39Which control BEST protects against unbounded consumption of an LLM API?
Question 40AI application logs are kept on the product team’s own server and nobody in the SOC sees them. What is the BEST improvement?
Question 41Which metric would BEST show senior management that AI security coverage is improving?
Question 42A provider announces it will retire the model version a company uses in 60 days. What should the security manager ensure FIRST?