練習できるAAISM問題には認定ガイド問題と解答とトレーニングを提供しています
無料ISACA AAISMテスト練習問題試験問題集
ISACA AAISM 認定試験の出題範囲：
| トピック | 出題範囲 |
|---|---|
| トピック 1 |
|
| トピック 2 |
|
| トピック 3 |
|
質問 # 214
Which of the following BEST ensures AI components are validated as part of disaster recovery testing?
- A. Simulating denial of service (DoS) attacks against AI APIs to evaluate detection capabilities
- B. Running simulated data loss scenarios by erasing test records from the AI system's feature store
- C. Disconnecting primary model training clusters to test retraining workflow during extended outages
- D. Monitoring model performance metrics during failover and recovery to assess system stability
正解：D
解説：
Business continuity and disaster recovery (BC/DR) exercises for AI must validate that critical AI components (feature stores, model registries, inference services, pipelines) operate within agreed recovery objectives during failover and restoration. Monitoring and evaluating model performance and stability during DR tests provides objective evidence that AI services remain functional, accurate, and reliable under contingency conditions, thereby validating the AI stack end-to-end.
Option A focuses on retraining during outages (a niche scenario) rather than validating service continuity for production inference. Option B is security testing, not BC/DR validation. Option C tests data loss handling but does not comprehensively validate AI service behavior across failover and recovery.
References: AI Security Management (AAISM) Body of Knowledge: "Operational Resilience-BC/DR for AI Systems," "Validation and Evidence of Continuity"; AAISM Study Guide: "AI DR Test Planning- Metrics, Model Performance Validation, and Recovery Readiness."
質問 # 215
The PRIMARY ethical concern of generative AI is that it may:
- A. Cause information to become unavailable
- B. Breach the confidentiality of information
- C. Cause information integrity issues
- D. Produce unexpected data that could lead to bias
正解：C
解説：
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
質問 # 216
An organization's CIO provided the AI steering committee with a list of AI technologies in use and tasked them with categorizing the technologies by risk. Which of the following should the committee do FIRST?
- A. Assess risk levels based on risk appetite and regulatory requirements
- B. Ensure the AI technologies are included in the asset inventory
- C. Begin grouping similar AI products and solutions together
- D. Identify vulnerabilities related to the technologies in use
正解：B
解説：
AAISM governance practices state that before categorizing technologies by risk, the first step is to ensure that all AI systems are documented in the organizational asset inventory. A complete inventory provides the foundation for subsequent risk analysis, accountability, and governance.
Grouping solutions, identifying vulnerabilities, and assessing risk levels come afterward, once inventory accuracy is established. Without confirming that the technologies are recorded in the inventory, risk categorization may miss critical assets.
質問 # 217
An AI application development team has been given access to user information and now must format it to be readable by the AI model. During which phase of the data life cycle would this MOST likely occur?
- A. Data normalization
- B. Data collection
- C. Data preparation
- D. Data minimization
正解：C
解説：
According to AAISM's data life-cycle model, data preparation is the phase where raw data is transformed into a model-ready format. The materials describe this phase as including "cleaning, encoding, formatting, feature engineering, and other transformations required for model consumption." This directly matches the scenario where a team formats user information to be readable by an AI model.
質問 # 218
Which of the following BEST helps an organization protect confidential data while using a pre- trained open-source AI solution?
- A. Periodically test the solution for vulnerabilities.
- B. Monitor the solution's performance for reliability.
- C. Continuously monitor usage to detect fraud and abuse.
- D. Verify and sanitize both inputs and outputs.
正解：D
解説：
Verifying and sanitizing inputs prevents sensitive data from being unintentionally exposed to the model, while sanitizing outputs helps ensure that confidential information is not leaked in responses, directly protecting data during use.
質問 # 219
Which of the following is the PRIMARY reason to use model cards for the management of AI systems?
- A. To standardize the model performance and intended use documentation.
- B. To facilitate the encryption of training data to prevent unauthorized model access.
- C. To help to ensure model retraining is automated and timely.
- D. To eliminate the need for human oversight of low-level decisions.
正解：A
解説：
Model cards are primarily used to provide standardized documentation about an AI model's intended use, performance characteristics, limitations, assumptions, risks, and evaluation results.
This supports transparency, governance, and informed decision-making throughout the AI system lifecycle.
質問 # 220
Which of the following BEST enables an organization to minimize risk from shadow AI?
- A. Prohibiting the use of any type of AI in the organization
- B. Implementing detective controls and training employees on the use of AI
- C. Implementing a robust framework for AI performance evaluation
- D. Prohibiting the use of shadow AI and establishing penalties for noncompliance
正解：B
解説：
Detective controls provide visibility into unauthorized AI usage, while employee training promotes awareness and responsible behavior. Together, they address both detection and prevention, making them the most effective approach to minimizing risks associated with shadow AI.
質問 # 221
An internet service provider (ISP) uses a chatbot based on a public model to help resolve connectivity issues. Customers report that the chatbot's instructions lack details about the ISP's proprietary equipment. Which of the following is the BEST way to improve the user experience?
- A. Use of additional public models
- B. Input filtering
- C. Output sanitization
- D. Retrieval-augmented generation (RAG)
正解：D
解説：
Retrieval-augmented generation integrates the chatbot with the organization's proprietary knowledge sources, enabling it to provide accurate and detailed responses specific to the ISP's equipment while maintaining the strengths of the base model.
質問 # 222
An organization is commissioning a third-party AI system using sensitive data. Which metric is MOST important to consider?
- A. Model response time
- B. Accuracy thresholds
- C. Service availability
- D. Accessibility rating
正解：B
解説：
AAISM specifies that when AI systems process sensitive or high-impact data (e.g., financial, identity, health), accuracy thresholds become critical risk indicators. Inaccurate predictions can cause harm, regulatory non- compliance, discrimination, or financial loss.
Availability (D) is important but secondary to correctness in sensitive decision scenarios. Response time (B) and accessibility ratings (A) do not outweigh the need for accurate, reliable predictions.
References: AAISM Study Guide - AI Model Risk Metrics; Accuracy Requirements for High-Impact Systems.
質問 # 223
Embedding unique identifiers into AI models would BEST help with:
- A. Tracking ownership
- B. Detecting adversarial attacks
- C. Eliminating AI system biases
- D. Preventing unauthorized access
正解：A
解説：
The AAISM framework explains that embedding unique identifiers--such as digital watermarks or model fingerprints--enables organizations to trace and verify model provenance. This technique is used for tracking ownership and intellectual property rights over models, particularly when sharing, licensing, or distributing AI systems. While identifiers may support certain security functions, their primary control objective is ownership verification, not preventing access, bias removal, or adversarial detection. The correct alignment with AAISM controls is tracking ownership.
質問 # 224
Which of the following AI data life cycle phases presents the GREATEST inherent risk?
- A. Training
- B. Preparation
- C. Monitoring
- D. Maintenance
正解：B
解説：
The data Preparation phase-covering sourcing, collection, labeling, cleansing, and provenance-presents the greatest inherent risk because it is where privacy, consent, representativeness, bias, quality, lineage, and legality must be established. Decisions and defects here propagate into training and downstream use, amplifying ethical, regulatory, and accuracy risks.
While Training can introduce additional risks (e.g., poisoning, leakage), these are frequently mitigated by controls that depend on having trustworthy prepared data. Monitoring and Maintenance are later-life-cycle phases oriented toward detection and correction; they are critical but inherently rely on the foundation set during Preparation.
References: AI Security Management™ (AAISM) Body of Knowledge: "AI Data Lifecycle Risks-Sourcing, Consent, Provenance, and Bias," "Risk Treatment Priorities Across the Lifecycle"; AAISM Study Guide:
"Data Preparation Controls and Quality Gates," "Bias and Privacy Risk Controls at Ingestion."
質問 # 225
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. Using training data from multiple sources
- B. Delivering AI-specific security awareness training
- C. Implementing an updated and tested break-glass policy
- D. Being transparent with customers about the data sources
正解：A
解説：
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.
質問 # 226
An organization is adopting an agentic AI solution from an external vendor to support internal IT operations. Which of the following provides the MOST reliable and independently verifiable evidence of implemented security controls?
- A. Internal red-team testing reports
- B. Third-party audit reports
- C. Industry benchmarking peer review
- D. General AI security whitepapers
正解：B
解説：
AAISM states that when evaluating external AI vendors, independently issued third-party audit reports (SOC, ISO, AI assurance assessments) provide the strongest evidence of implemented controls because they are objective, repeatable, and externally verified.
質問 # 227
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?
- A. Rely primarily on vendor-provided security features and seek third-party certifications
- B. Apply AI-specific controls from external frameworks without customization and initiate monitoring to expedite compliance
- C. Postpone control selection until deployment and address risk through enhanced monitoring
- D. Map identified AI threats to enterprise control catalogs and integrate AI-specific safeguards where gaps exist
正解：D
解説：
AAISM requires that control selection be threat-led and context-specific, aligning AI threats to the organization's existing enterprise control catalogs (security, privacy, resilience) and augmenting them with AI-specific safeguards where coverage is insufficient. This ensures consistency with the risk appetite, removes duplication, and closes AI-unique gaps (e.g., prompt injection, data leakage from context windows, model misuse). Generic reliance on vendors or uncustomized external frameworks does not ensure fit-for-purpose coverage, and deferring control selection to post-deployment contradicts proactive risk treatment.
質問 # 228
Which of the following security framework elements BEST helps to safeguard the integrity of outputs generated by AI algorithms?
- A. Management is prepared to disclose AI system architecture to stakeholders
- B. Risk exposure due to bias in AI outputs is kept within an acceptable range
- C. Responsibility is defined for legal actions related to AI regulatory requirements
- D. Ethical standards are incorporated into security awareness programs
正解：B
解説：
According to AAISM technical controls, the element of security frameworks that directly safeguards output integrity is ensuring that bias-related risks are maintained within acceptable ranges. This ensures that AI results remain consistent, fair, and reliable, preserving trust in outputs. Ethical awareness, architectural disclosure, and legal responsibilities are governance practices but do not directly secure output integrity.
Output integrity is primarily protected through bias management and ongoing evaluation of fairness and accuracy.
References:
AAISM Exam Content Outline - AI Technologies and Controls (Integrity of AI Outputs) AI Security Management Study Guide - Bias and Output Integrity Controls
質問 # 229
When addressing privacy concerns related to AI systems, which of the following is the GREATEST significance of user consent for an organization?
- A. It helps the organization detect biases and ensure fairness
- B. It allows the organization to process user data in the AI system
- C. It prevents unauthorized access to data within the AI system
- D. It enables users to delete and modify their personal data
正解：B
解説：
Within AAISM's privacy governance, consent is a lawful basis that authorizes processing of personal data for defined purposes. Its principal significance is granting the organization the authority to process user data in AI workflows in line with stated purposes and limits. While fairness (A) and security controls (C) are essential, they are distinct obligations; data subject rights such as rectification/erasure (B) exist regardless of consent and are not "enabled" by it. Therefore, the greatest significance of consent is that it legally permits processing under declared purposes and constraints.
References: AI Security Management (AAISM) Body of Knowledge - Privacy & Data Protection in AI; Lawful Basis and Purpose Limitation. AAISM Study Guide - Consent, Transparency, and Data Subject Rights; Privacy-by-Design Controls in AI Pipelines.
質問 # 230
Which of the following would BEST protect an AI system against an evasion attack?
- A. Validation of training datasets
- B. Robust access controls
- C. Adversarial training
- D. Context awareness training
正解：C
解説：
Adversarial training exposes the model to manipulated or adversarial inputs during training, enabling it to recognize and resist such inputs during operation, which directly mitigates evasion attacks.
質問 # 231
Which of the following is the GREATEST benefit of performing AI security risk assessments?
- A. Appropriate privacy risk controls are implemented for AI models
- B. Risk prioritization decisions are made for AI security
- C. The appropriate level of funding is secured for AI security risk
- D. The risk register is updated with the latest AI risk
正解：B
解説：
AAISM emphasizes that the core outcome of AI risk assessments is prioritization: mapping threat likelihood and business impact to determine which risks to treat first, at what strength, and with which controls.
Implementing privacy controls (A), funding alignment (B), and updating registers (C) are important outputs, but the greatest benefit is making defensible, prioritized decisions that align with risk appetite and optimize control selection and resource allocation.
References: AI Security Management™ (AAISM) Body of Knowledge - AI Risk Assessment & Treatment; Risk Appetite, Tolerance, and Prioritization; Governance of Risk Decisions and Tracking.
質問 # 232
A CISO has been tasked with providing key performance indicators (KPIs) on the organization's newly launched AI chatbot. Which of the following are the BEST metrics for the CISO to recommend?
- A. Customer effort score and user retention rate
- B. Error rate and bias detection
- C. Response time and throughput
- D. Explainability and F1 score
正解：B
解説：
For executive security and governance reporting, AAISM prioritizes risk- and harm-oriented KPIs that reflect safety, reliability, and responsible behavior of AI systems. Error rate (safety/quality signal) and bias detection (fairness/compliance signal) provide leading indicators of model risk, potential user harm, and regulatory exposure-key interests for a CISO. Explainability and F1 (A) are model performance/interpretability metrics; customer effort/retention (B) are business CX metrics; response time/throughput (C) are operational SRE metrics. While valuable, they are secondary to risk-centric KPIs for CISO oversight.
References: AI Security Management™ (AAISM) Body of Knowledge - AI Risk Metrics and Assurance; Governance Dashboards for AI. AAISM Study Guide - Operationalizing AI Controls; Safety, Fairness, and Compliance Indicators for Executive Reporting.
O Error rate and bias detection
質問 # 233
A retailer's AI-driven dynamic pricing system failed after its feature store was corrupted, resulting in all prices being set to zero. The data source has been isolated, and management demands immediate recovery. Which of the following is the MOST appropriate course of action to restore operations while minimizing impact?
- A. Failover to a secondary environment and retrain the model with recent data.
- B. Require customer service to manually price and approve orders until the AI system is fully restored.
- C. Revert to the previous AI model and validate backup of the feature store.
- D. Keep the system offline and conduct a full forensic investigation.
正解：C
解説：
Reverting to a previously stable model and validating a clean backup of the feature store enables rapid restoration of normal operations while ensuring data integrity. This minimizes business disruption and avoids propagating corrupted data into the system.
質問 # 234
An AI system that supports critical processes has deviated from expected performance and is producing biased outcomes. Which of the following is the BEST course of action?
- A. Retrain the model with a new and expanded dataset
- B. Activate the model kill switch
- C. Conduct audits of the data and the model
- D. Perform a root cause analysis to identify mitigation steps
正解：D
解説：
AAISM directs that when harmful or biased behavior is observed in a production AI system, the organization should enter a formal incident/variance handling workflow that begins with root cause analysis (RCA) to identify the source of deviation (data drift, concept drift, feature leakage, pipeline changes, control failures) and determine proportionate risk treatments. Immediate retraining (Option A) without RCA risks reinforcing the same bias; audits (Option C) are key activities within RCA rather than the action that frames the response; a kill switch (Option D) is reserved for conditions where risk exceeds the defined tolerances and immediate harm prevention is required.
References: AI Security Management (AAISM) Body of Knowledge - Incident Response & Post-Incident Improvement; Model Risk Treatment & Drift Management; Bias Detection and Remediation Governance.
質問 # 235
Which of the following is the GREATEST concern when a vendor enables generative AI features for an organization's critical system?
- A. Proposed regulatory enhancements
- B. Security monitoring and alerting
- C. Access to the model
- D. Bias and ethical practices
正解：C
解説：
When enabling genAI capabilities in a critical system, AAISM prioritizes controlling access to the model and its interfaces (prompt surfaces, context windows, tools/functions, and connected data) because exposure expands the attack surface for prompt injection, data exfiltration, jailbreaks, and misuse.
質問 # 236
An organization plans to leverage AI in the software development process to speed up coding. Which of the following should the information security manager do FIRST?
- A. Train developers to verify AI output
- B. Perform a cost-benefit analysis
- C. Update the security policy to include AI controls
- D. Conduct an impact assessment
正解：D
解説：
AAISM guidance specifies that before introducing AI into any business or technical workflow, an AI Impact Assessment must be conducted early to determine potential risks, privacy implications, misuse scenarios, governance gaps, and required security controls. This aligns with the principle that AI adoption must begin with governance and risk identification, not training or policy modification.
Training developers (B) is important but occurs after identifying risks. Updating policies (C) is also downstream of the assessment. Cost-benefit analysis (D) supports business justification but does not address security.
References: AAISM Study Guide - AI Governance; Impact Assessment Required Before Deployment.
質問 # 237
Which of the following controls BEST mitigates the risk of data poisoning?
- A. Digital watermarking
- B. Data validation
- C. Intrusion detection
- D. Data set restoration
正解：B
解説：
The AAISM technical controls framework emphasizes data validation as the primary safeguard against data poisoning attacks. Poisoning occurs when attackers insert malicious or corrupted data into training sets. Validation techniques verify the quality, authenticity, and consistency of input data before training, preventing compromised samples from corrupting the model.
Restoration helps after compromise, watermarking protects ownership, and intrusion detection monitors networks rather than data quality. The most effective preventive measure is data validation.
質問 # 238
Which of the following key risk indicators (KRIs) is MOST relevant when evaluating the effectiveness of an organization's AI risk management program?
- A. Percentage of AI projects in compliance
- B. Number of AI models deployed into production
- C. Number of AI-related training requests submitted
- D. Percentage of critical business systems with AI components
正解：A
解説：
AAISM identifies percentage of AI projects in compliance as the most relevant KRI for evaluating AI risk management effectiveness. This metric directly reflects adherence to governance, regulatory, and security requirements. The number of models deployed (A) or systems with AI components (B) indicate scale, not risk management quality. Training requests (D) show awareness levels but do not measure effectiveness of risk management. Compliance percentage provides a direct, measurable indication of how well risks are being governed and mitigated.
質問 # 239
......
試験準備には欠かさない！トップクラスのISACA AAISM試験アプリ学習ガイドで練習問題最新版：https://www.jpntest.com/shiken/AAISM-mondaishu
問題集練習試験問題学習ガイドはAAISM試験：https://drive.google.com/open?id=1i3JQJ38qDR5zMpBZ4Uf64hTAsw5GF7xH