Isaca AAISM Online Practice Questions and Exam Preparation

Isaca AAISM Online Questions & Answers

• Question 71:

  A financial organization uses AI to detect potential fraudulent activities but is concerned about the impact of potential data poisoning. Which of the following controls would BEST mitigate this risk?

  A. Being transparent with customers about the data sources
  B. Implementing an updated and tested break-glass policy
  C. Delivering AI-specific security awareness training
  D. Using training data from multiple sources
  D. Using training data from multiple sources

  ---

  AAISM identifies training-data diversity and provenance assurance as primary treatments against data poisoning. Sourcing data from multiple, independently governed providers, combined with ingestion validation and anomaly screening, reduces the chance that a single compromised source can skew model behavior and improves cross-source consistency checks. Transparency, break-glass, and awareness are valuable but do not directly reduce poisoning exposure at the training boundary.

  References: AI Security Management. (AAISM) Body of Knowledge - Data Governance and Integrity for AI; Adversarial ML: Poisoning Threats and Mitigations; Supplier and Source Diversification Controls.

• Question 72:

  Which of the following technologies can be used to manage deepfake risk?

  A. Systematic data tagging
  B. Multi-factor authentication (MFA)
  C. Blockchain
  D. Adaptive authentication
  C. Blockchain

  ---

  The AAISM study material highlights blockchain as a control mechanism for managing deepfake risk because it provides immutable verification of digital media provenance. By anchoring original data signatures on a blockchain, organizations can verify authenticity and detect tampered or synthetic content. Data tagging helps organize but does not guarantee authenticity. MFA and adaptive authentication strengthen identity security but do not address content manipulation risks. Blockchain's immutability and

  traceability make it the recognized technology for mitigating deepfake challenges.

  References:

  AAISM Study Guide - AI Technologies and Controls (Emerging Controls for Content Authenticity)

  ISACA AI Governance Guidance - Blockchain for Data Integrity and Deepfake Mitigation

• Question 73:

  An organization uses an AI tool to scan social media for product reviews. Fraudulent social media accounts begin posting negative reviews attacking the organization's product. Which type of AI attack is MOST likely to have occurred?

  A. Model inversion
  B. Deepfake
  C. Availability attack
  D. Data poisoning
  C. Availability attack

  ---

  The AAISM materials classify availability attacks as attempts to disrupt or degrade the functioning of an AI system so that its outputs become unreliable or unusable. In this scenario, the fraudulent social media accounts are deliberately overwhelming the AI tool with misleading negative reviews, undermining its ability to deliver accurate sentiment analysis. This aligns directly with the concept of an availability attack. Model inversion relates to reconstructing training data from outputs, deepfakes involve

  synthetic content generation, and data poisoning corrupts the training set rather than manipulating inputs at runtime. Therefore, the fraudulent review campaign is most accurately identified as an availability attack.

  References:

  AAISM Study Guide - AI Risk Management (Adversarial Threats and Availability Risks)

  ISACA AI Security Management - Attack Classifications

• Question 74:

  The PRIMARY ethical concern of generative AI is that it may:

  A. Produce unexpected data that could lead to bias
  B. Cause information integrity issues
  C. Cause information to become unavailable
  D. Breach the confidentiality of information
  B. Cause information integrity issues

  ---

  AAISM materials emphasize that the primary ethical concern with generative AI is the risk to information integrity. Generative models can create content that appears authentic but is fabricated, misleading, or manipulated. This undermines trust in information ecosystems and can have wide-reaching social, legal, and organizational impacts. While confidentiality breaches and bias are concerns, they are not the central ethical issue inherent to generative models. Availability is less relevant in this context. The

  most pressing concern is that generative AI may compromise the integrity of information.

  References:

  AAISM Study Guide - AI Risk Management (Ethical Risks of Generative AI)

  ISACA AI Security Management - Integrity Concerns in Generative Systems

• Question 75:

  Which of the following actions BEST enables the evaluation of bias during an AI impact assessment?

  A. Assessing the AI system's training data to ensure it represents all relevant end-user groups
  B. Comparing the AI system's output against historical data benchmarks
  C. Analyzing the AI system's reaction time under peak workload conditions
  D. Measuring the AI system's performance processing speed under predefined varying workloads
  A. Assessing the AI system's training data to ensure it represents all relevant end-user groups

  ---

  In AI Security Management, bias evaluation is primarily a data problem before it is a model or performance problem. The official AI Security Management content explains that impact assessments must specifically analyze training data representativeness against the demographics and characteristics of all relevant users and stakeholders. If the data used to train an AI system underrepresents or omits particular groups, the resulting model will systematically disadvantage them, regardless of how fast or "accurate" it is on average. While comparing outputs to historical data (option B) can help, historical data itself may be biased. Performance-related options (C and D) relate to efficiency and scalability, not fairness. Therefore, systematically assessing whether the training data covers all relevant end-user groups is the most direct and effective way to evaluate and mitigate bias.

  References: AI Security Management (AAISM) Study Guide - AI Risk Identification and Impact Assessment; Data Governance and Bias Section.

• Question 76:

  Security and assurance requirements for AI systems should FIRST be embedded in the:

  A. Model design phase
  B. Model training phase
  C. Model testing phase
  D. Model deployment phase
  A. Model design phase

  ---

  AAISM directs organizations to embed security, safety, and compliance controls at design time ("secure-by-design" and "shift-left"), ensuring requirements for robustness, privacy, and governance are defined as non-functional constraints on architecture, data sourcing, model choices, and evaluation criteria before any model is trained. Deferring these requirements to training, testing, or deployment increases residual risk and rework, and weakens traceability of control coverage. References:. AI Security Management (AAISM) Body of Knowledge: Governance-Secure-by-Design; Policy-to-Control Traceability; Requirements Management. AAISM Study Guide: AI Program Lifecycle-Planning and Design Controls; Design-time Threat Modeling and Control Selection. AAISM Mapping to Standards: Design-phase Risk Identification and Requirements Engineering for AI

• Question 77:

  When implementing a generative AI system, which of the following approaches will BEST prevent misalignment between the corporate risk appetite and tolerance?

  A. Ensuring effective AI key performance indicators (KPIs)
  B. Performing an AI impact assessment
  C. Creating and maintaining an AI risk register
  D. Establishing and monitoring acceptable levels of AI system risk
  D. Establishing and monitoring acceptable levels of AI system risk

  ---

  AAISM governance guidance specifies that alignment between AI system adoption and organizational risk appetite is achieved by defining and monitoring acceptable levels of system risk. This ensures that generative AI operations remain within boundaries approved by leadership and compliance frameworks. While KPIs track performance, they do not ensure alignment with risk tolerance. AI impact assessments help identify risks but do not maintain continuous oversight. A risk register records risks but does

  not dynamically enforce acceptable thresholds. The most effective governance approach is to establish and monitor acceptable AI system risk levels.

  References:

  AAISM Study Guide - AI Governance and Program Management (Risk Appetite and Tolerance Alignment)

  ISACA AI Security Management - Generative AI Risk Governance

• Question 78:

  Which of the following BEST describes an adversarial attack on an AI model?

  A. Attacking underlying hardware
  B. Providing inputs that mislead the model into incorrect predictions
  C. Reverse-engineering the model using social engineering
  D. Conducting denial-of-service attacks on AI APIs
  B. Providing inputs that mislead the model into incorrect predictions

  ---

  AAISM defines adversarial attacks as manipulations of input data (text, image, audio, numeric values) designed to cause the model to produce incorrect or harmful predictions. Hardware attacks (A) are infrastructure threats. Social engineering (C) targets people, not models. DoS attacks (D) affect availability, not model decision pathways.

  References: AAISM Study Guide - Adversarial Threats; Input Manipulation.

• Question 79:

  Which of the following mitigation control strategies would BEST reduce the risk of introducing hidden backdoors during model fine-tuning via third-party components?

  A. Leveraging open-source models and packages
  B. Performing threat modeling and integrity checks
  C. Disabling runtime logs during model training
  D. Implementing unsupervised learning methods
  B. Performing threat modeling and integrity checks

  ---

  AAISM highlights threat modeling and supply chain integrity checks as key controls for managing AI-specific risks, including hidden backdoors in third-party models, libraries, or fine-tuning artifacts. The official guidance states that organizations should "identify adversarial insertion points, verify component integrity, and continuously test for malicious behaviors introduced via external components." This is precisely what option B describes. Simply using open-source (A) does not guarantee security; code can still contain malicious modifications. Disabling runtime logs (C) would reduce visibility, making backdoor detection harder. Choosing unsupervised learning (D) is a modeling approach and has no inherent relation to backdoor risk reduction. The explicit combination of AI-focused threat modeling and integrity verification of external components is the recommended best practice to mitigate this class of attack.

  References: AI Security Management (AAISM) Study Guide - AI Supply Chain Risk Management; Threat Modeling and Component Integrity.

• Question 80:

  When addressing privacy concerns related to AI, what is the GREATEST significance of user consent?

  A. It prevents unauthorized access to data
  B. It enables deletion/modification of personal data
  C. It allows the organization to process user data in the AI system
  D. It helps detect bias and ensure fairness
  C. It allows the organization to process user data in the AI system

  ---

  AAISM clarifies that the primary regulatory purpose of user consent is to provide lawful authorization for processing personal data, including use in AI models.

  Consent does not directly prevent unauthorized access (A). Deletion rights (B) relate to data subject rights but are not the purpose of consent. Bias detection (D) is unrelated.

  References: AAISM Study Guide - AI Privacy and Legal Basis for Processing Data.