AAISM Practice Test 19 — 15 Questions

Question 1

An AI risk manager at an insurance company is designing controls for a claims-triage model that recommends whether a claim should be fast-tracked or manually reviewed. The model's recommendations influence but do not determine the final decision. The manager wants to ensure that claimants who are adversely affected can understand and contest the outcome. Which control BEST addresses this requirement?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Provide claimants with a plain-language explanation of the factors influencing the recommendation and a documented route to request human review.. The requirement is that adversely affected claimants can understand and contest outcomes. A plain-language explanation plus a documented human review route delivers both transparency and contestability. Threshold tuning, audit logging, and retraining improve performance or oversight but do not provide the affected individual with an explanation or a remedy.

Answer

Log all model inputs and outputs to an immutable audit store for later forensic analysis by the compliance team.

Answer

Increase the model's confidence threshold so that only high-certainty claims are fast-tracked.

Answer

Retrain the model quarterly on the most recent claims data to reduce the frequency of adverse recommendations.

Question 2

An AI risk manager at a telecommunications company is establishing an ongoing monitoring program for a network-fault prediction model deployed in production. The model triggers field technician dispatches. The manager wants to define monitoring signals that would indicate the model's risk profile has materially changed and require re-assessment. Which TWO signals are MOST appropriate to include? (Choose two.)

Accepted Answer

A measurable decline in prediction accuracy or an increase in false dispatch rates relative to validated thresholds.. Ongoing model risk monitoring should track both input distributions and outcome performance. Feature drift against the training baseline reveals that production conditions have changed, while accuracy and false dispatch rates show whether the model still performs as validated. Code volume, team tenure, and latency are operational or organizational metrics that do not by themselves indicate a material change in model risk.

Answer

The number of lines of code in the model's serving application increasing between releases.

Answer

An increase in the model's inference latency measured in milliseconds per prediction.

Answer

The average tenure of the data science team maintaining the model decreasing over time.

Question 3

A telecommunications provider has completed its first AI governance maturity assessment. Leadership wants to move from ad hoc AI oversight to a repeatable, enterprise-wide program that can demonstrate continuous improvement to the board. Which action best advances the program toward that goal?

Accepted Answer

Adopting a recognized AI governance framework and mapping current practices to its domains to identify and close gaps. Moving from ad hoc oversight to a repeatable program requires a structured reference model. Adopting a recognized AI governance framework and mapping current practices to its domains lets the organization benchmark maturity, prioritize gaps, and report measurable progress to the board, which is the essence of continuous improvement.

Answer

Hiring a Chief AI Officer to personally approve every AI use case before deployment

Answer

Purchasing an AI governance platform to centralize policy documents and workflow approvals

Answer

Commissioning an external legal review of all AI regulations applicable to the telecommunications sector

Question 4

A security team is deploying a fraud detection model that processes real-time transaction data. The model was trained on historical data containing sensitive customer information. The team wants to ensure that individual customer records cannot be reconstructed from the model's outputs or parameters. Which technique should they implement?

Accepted Answer

Differential privacy during training. Differential privacy is the only technique listed that provides a formal guarantee limiting the influence of any single training record on the model. By adding noise during training, it directly prevents reconstruction and membership inference attacks, which is exactly what the team needs to protect sensitive customer data in a fraud detection model.

Answer

Secure multiparty computation for model inference

Answer

Federated learning across customer devices

Answer

Homomorphic encryption of the training data

Question 5

A global manufacturer is standing up an AI governance program and must assign accountability for AI risks across the enterprise. The Chief Risk Officer wants to ensure that accountability is clearly allocated for both model development and ongoing operation. Which TWO structural elements are most essential to establish this accountability? (Choose two.)

Accepted Answer

A RACI matrix that assigns AI risk ownership to named roles across the lifecycle. A RACI matrix assigning AI risk ownership to named roles across the lifecycle and a board-approved governance charter defining decision rights and escalation paths together establish clear, enforceable accountability. The matrix provides role-level ownership for development and operations, while the charter provides formal authority and escalation, ensuring no gaps remain.

Answer

A centralized AI ethics committee that reviews all AI projects before launch

Answer

A quarterly AI performance dashboard presented to the executive committee

Answer

A vendor risk assessment template for third-party AI components

Question 6

A multinational financial services firm has deployed a generative AI assistant for its customer support agents. The AI governance committee has approved the tool, but the Chief Information Security Officer is concerned that the firm cannot demonstrate to regulators that the AI's outputs are being monitored for harmful or biased content on an ongoing basis. Which of the following is the MOST appropriate control to address this concern?

Accepted Answer

Implement an AI output monitoring and logging solution that captures model responses, flags anomalous or biased outputs, and retains records for audit review.. The correct answer is the implementation of an AI output monitoring and logging solution. This provides the continuous, auditable evidence regulators expect, showing that the organization actively monitors for harmful or biased content. Manual review, one-time assessments, and template restrictions do not deliver the systematic oversight and audit trail required for AI governance.

Answer

Require customer support agents to manually review every AI-generated response before sending it to customers.

Answer

Disable the AI assistant's ability to generate free-text responses and restrict it to predefined templates.

Answer

Conduct a one-time bias assessment of the AI model before deployment and document the results.

Question 7

A security engineer is evaluating an AI-based fraud detection system that uses a gradient boosting model. The engineer wants to test the model's robustness against an attacker who can only query the model through a public API and observe the predicted labels and confidence scores. The attacker has no access to the model's architecture or training data. Which attack category best describes this scenario?

Accepted Answer

Black-box evasion attack. The scenario describes an attacker with only query access to a deployed model, aiming to cause misclassification. This is a classic black-box evasion attack. White-box attacks require internal model access, poisoning affects training, and model inversion extracts data rather than evading. Thus, the correct classification is black-box evasion.

Answer

Model inversion attack

Answer

Data poisoning attack

Answer

White-box evasion attack

Question 8

A healthcare AI governance committee is reviewing a sepsis prediction model deployed across three hospitals. The model was trained on data from one hospital and shows a 12% drop in sensitivity when used at a second hospital with a different patient demographic. The committee must decide how to treat this risk. Which action is MOST appropriate?

Accepted Answer

Apply site-specific thresholds and enhanced monitoring at the second hospital while planning retraining.. The risk is a site-specific performance disparity, not a global model failure. The most appropriate treatment is to mitigate at the affected site through threshold adjustment and monitoring while addressing the root cause through retraining with more representative data. This preserves benefit where the model is validated and manages harm where it is not.

Answer

Replace the model with a simple rule-based sepsis score at all hospitals.

Answer

Suspend the model at all hospitals until it is retrained on data from every hospital.

Answer

Document the performance drop as a known limitation and continue use without changes.

Question 9

A multinational corporation is establishing an AI governance board. The board's charter must define its responsibilities for overseeing AI risks across the enterprise. Which TWO responsibilities are most critical for the board to own? (Choose two.)

Accepted Answer

Ensuring that AI risk management processes are integrated into enterprise risk management.. The governance board's most critical responsibilities are setting the AI risk appetite and ensuring AI risk management is integrated into enterprise risk management. These activities provide strategic direction and embed AI risk oversight into the broader organizational risk framework, which are essential for effective governance. Technical and operational tasks are delegated to management and staff.

Answer

Conducting individual performance reviews of AI developers.

Answer

Writing the code for AI risk mitigation controls.

Answer

Reviewing and approving the technical architecture of each AI model.

Question 10

A global bank's AI governance program is being audited. The auditors find that the bank has a robust AI risk assessment process for new models, but no mechanism to monitor risks that emerge after deployment, such as data drift or changing regulatory requirements. Which control should the governance team implement to address this gap?

Accepted Answer

A periodic AI model re-validation and risk reassessment process tied to a defined schedule and trigger events.. The audit gap is the absence of post-deployment risk monitoring. A periodic re-validation and reassessment process, scheduled and event-triggered, ensures that models are continually evaluated against performance, data drift, and regulatory changes. This control provides the necessary oversight to maintain risk management throughout the AI lifecycle and satisfies the auditors' concern about ongoing risk identification.

Answer

A requirement for business units to self-certify that their AI models remain compliant.

Answer

A one-time review of all deployed models at the end of the fiscal year.

Answer

An annual penetration test of the AI infrastructure.

Question 11

An AI risk manager is preparing a board-level report on a portfolio of machine learning models. The board wants to understand which models pose the greatest potential for financial loss. Which approach BEST aligns model risk assessment with enterprise risk management?

Accepted Answer

Score each model by combining decision impact, failure likelihood, and control effectiveness into a risk rating.. Aligning model risk with enterprise risk management requires expressing exposure in terms of impact, likelihood, and control effectiveness. That structure allows the board to compare model risk against other enterprise risks and to prioritize oversight where potential financial loss is greatest. Inventory and vulnerability counts alone do not convey that exposure.

Answer

List all models with their accuracy metrics and let the board interpret the financial implications.

Answer

Report the total number of models in production and the growth rate of the AI portfolio.

Answer

Rank models by the number of open security vulnerabilities reported by the scanning tool.

Question 12

A healthcare organization is using an AI model to predict patient readmission risk. The model was trained on historical data that includes a feature indicating whether the patient was enrolled in a chronic care management program. The security team is concerned that the model may inadvertently discriminate against a protected group. Which technique should they use to detect and mitigate bias in the model's predictions?

Accepted Answer

Adversarial debiasing. Adversarial debiasing is designed to reduce bias by training an adversary to predict the protected attribute and penalizing the model if it can. This directly mitigates discrimination. Data augmentation, pruning, and cross-validation do not specifically address fairness concerns.

Answer

Cross-validation

Answer

Model pruning

Answer

Data augmentation

Question 13

A retail company's AI governance policy requires that all AI systems undergo a risk assessment before deployment. A business unit wants to deploy a chatbot for customer service that uses a large language model (LLM) from a public API. The business unit argues that because the LLM is pre-trained and hosted externally, a full risk assessment is unnecessary. What should the governance team do?

Accepted Answer

Require the business unit to conduct a risk assessment focused on the specific use case and data flows.. The governance team should require a risk assessment tailored to the specific use case and data flows. This ensures that risks unique to the integration—such as data privacy, model bias, and output accuracy—are evaluated and managed. It upholds the governance policy while allowing the business to proceed responsibly, balancing innovation with oversight.

Answer

Approve the deployment immediately because the LLM provider is responsible for all risks.

Answer

Waive the risk assessment because the LLM is a third-party service.

Answer

Prohibit the use of public LLM APIs entirely.

Question 14

A security analyst is reviewing the risks associated with using a third-party pre-trained model for a new AI application. The analyst must identify TWO risks that are specifically related to the model's provenance and supply chain. (Choose two.)

Accepted Answer

The model may contain a backdoor that triggers malicious behavior on specific inputs.. The two supply chain risks are a hidden backdoor and licensing violations. A backdoor is an intentional malicious modification, while licensing issues arise from improper data usage during training. Both stem from a lack of trust in the model's provenance. Bias, explainability, and resource requirements are not specific to supply chain.

Answer

The model may require more computational resources than available, causing performance issues.

Answer

The model's architecture may be too complex to interpret, leading to a lack of explainability.

Answer

The model may have been trained on data that includes biased or unrepresentative samples.

Question 15

An AI security analyst is reviewing a system that uses a large language model (LLM) for customer support. The LLM is fine-tuned on internal documents and deployed via an API. The analyst wants to detect if an attacker is attempting to extract sensitive training data through crafted prompts. Which control is most effective for this detection?

Accepted Answer

Analyzing query patterns for repeated or systematic probing of model outputs. Detecting training data extraction requires identifying the behavioral signature of an attacker systematically probing the model. Monitoring query patterns for repetition, structure, and volume can reveal such attempts, whereas watermarking, similarity checks, and injection firewalls address different threats and do not reliably detect extraction.

Answer

Monitoring for unusually high similarity between prompts and training data

Answer

Deploying a prompt injection firewall that filters malicious inputs

Answer

Implementing output watermarking to trace generated text