AAISM Practice Test 17 — 15 Questions

Question 1

What is the purpose of 'AI Model Transparency' in a risk management framework?

Accepted Answer

To enable stakeholders to evaluate model safety and performance.. AI Model Transparency enables stakeholders — auditors, regulators, business owners, and affected users — to understand how a model behaves, what data it uses, and how it performs, so they can evaluate its safety, fairness, and fitness for purpose. It is a governance enabler, not a performance optimization.

Answer

To speed up the training process.

Answer

To hide the model architecture from competitors.

Answer

To minimize the number of API calls.

Question 2

What is the primary function of an 'AI Steering Committee'?

Accepted Answer

Strategic oversight, prioritization, and risk management.. An AI Steering Committee's primary function is strategic oversight, prioritization, and risk management for AI initiatives across the organization. It sets direction, approves high-risk use cases, allocates resources, and ensures AI activities align with business strategy and risk appetite. It is a governance body, not an operational or technical execution team.

Answer

Conducting individual performance reviews.

Answer

Writing all the software code.

Answer

To manage daily printer issues.

Question 3

When drafting an AI policy, how should the organization address 'Human-in-the-Loop' (HITL) requirements?

Accepted Answer

Mandate HITL for high-impact or sensitive decision-making processes.. HITL should be applied proportionally to risk, so the policy must mandate human review specifically for high-impact or sensitive decisions such as credit, hiring, medical, or legal outcomes. Requiring HITL for every inference is operationally impossible at scale and defeats the purpose of automation. The correct answer reflects a risk-based governance approach where oversight intensity matches the potential harm of the AI decision.

Answer

Require HITL for every single AI inference.

Answer

Delegate HITL requirements to the software development team.

Answer

Exclude HITL to maximize system throughput.

Question 4

How should an organization classify AI systems for governance?

Accepted Answer

By the potential impact on safety, privacy, and business operations.. AI governance classification must be risk-based, driven by the potential impact an AI system can have on safety, privacy, and business operations. This aligns with frameworks like the NIST AI RMF and EU AI Act, which tier systems by risk level (unacceptable, high, limited, minimal) to determine the controls, oversight, and documentation required. Technical implementation details such as language or team size do not determine governance obligations.

Answer

By the programming language used.

Answer

By the number of developers assigned.

Answer

By the date of deployment.

Question 5

You are using the Adversarial Robustness Toolbox (ART) to assess a classifier. Which attack technique would you simulate to test the model's vulnerability to 'Evasion' attacks?

Accepted Answer

Fast Gradient Sign Method (FGSM). The Fast Gradient Sign Method (FGSM) is a well-known evasion attack that generates adversarial examples by adding small perturbations to input data to fool a classifier. It is specifically designed to test a model's vulnerability to evasion, where the attacker manipulates inputs at inference time to cause misclassification. Therefore, simulating FGSM with ART is the correct choice for assessing evasion vulnerabilities.

Answer

Membership Inference

Answer

Model Inversion

Answer

Data Poisoning via label flipping

Question 6

Which THREE metrics should be monitored for 'Model Performance' and security?

Accepted Answer

Inference latency.. Inference latency (A) is a core Model Performance metric because it measures the time from request to prediction, and a rising latency can indicate resource contention, model degradation, or a denial-of-service condition affecting availability. Unexpected error rate increases (B) are essential for both performance and security, since spikes in failed inferences, exceptions, or 5xx responses can reveal model bugs, bad inputs, or adversarial probing and attacks. Model drift (D), the decay of predictive performance over time as data distributions change, must be monitored to detect accuracy degradation and potential data poisoning or concept drift that undermines model reliability. The unmarked options do not belong: developer coffee breaks (C) are irrelevant to system metrics, and total electricity cost (E) is a facilities or billing concern, not a model performance or security indicator.

Answer

Number of coffee breaks taken by developers.

Answer

Total cost of electricity.

Question 7

When securing an AI-driven chatbot, which THREE of the following are recommended practices to mitigate 'Prompt Injection' vulnerabilities?

Accepted Answer

Perform input sanitization on all user-submitted text. Input sanitization (A) is correct because stripping or escaping control characters, delimiters, and instruction-like patterns from user-submitted text reduces the attacker's ability to smuggle directives that override the intended prompt. Using system messages to enforce strict context boundaries (B) is correct because a privileged system prompt that defines the assistant's role, allowed scope, and instruction hierarchy helps the model distinguish trusted developer instructions from untrusted user content. Deploying a refusal layer for unexpected system-level queries (D) is correct because detecting and rejecting requests that attempt to access system prompts, tools, or internal configuration blocks escalation paths typical of prompt injection. Option C is wrong because storing training data in plaintext increases exposure and does nothing to mitigate prompt injection. Option E is wrong because raising the temperature parameter increases randomness and creativity, making outputs less predictable and potentially easier to manipulate, rather than hardening the chatbot against injection.

Answer

Store model training data in plaintext

Answer

Increase the model's temperature parameter

Question 8

Which TWO of the following are risks associated with using public pre-trained models without further security validation?

Accepted Answer

Hidden backdoors injected into the model weights. Option D is correct because public pre-trained models can contain hidden backdoors or maliciously poisoned weights, so deploying them without security validation may let an attacker trigger unintended behavior at inference time. Option E is correct because such models are often trained on datasets with unclear or unlicensed provenance, creating legal and compliance exposure such as copyright or data-usage violations. Option A is not a risk but a potential benefit, since public models are not inherently more interpretable and may actually be less transparent. Option B is not a risk but a performance claim, and public models do not automatically reduce inference latency versus custom models. Option C is also not a risk, as hardware acceleration depends on the runtime and framework, not on the model being public or pre-trained.

Answer

Improved model interpretability

Answer

Reduced inference latency compared to custom models

Answer

Automatic hardware acceleration

Question 9

Which TWO aspects of 'Transparency' are most important for AI stakeholders?

Accepted Answer

Disclosing the use of AI to the end-user.. Transparency for AI stakeholders centers on making the AI's role and reasoning visible and understandable. Option C is correct because disclosing the use of AI to the end-user ensures people know when they are interacting with or affected by an AI system, which is a foundational transparency requirement in AI ethics frameworks and regulations such as the EU AI Act's transparency obligations. Option D is correct because providing a clear explanation of how the AI reaches decisions (e.g., explainability, interpretability, or model cards) lets stakeholders understand, audit, and challenge outputs, which is essential for accountability and trust. Option A is not relevant because server rack IDs are infrastructure asset details unrelated to AI transparency toward stakeholders. Option B is not relevant because marketing budget information does not reveal how the AI works or that it is being used. Option E is not relevant and would violate privacy, as developer home addresses have no bearing on AI transparency.

Answer

Providing a list of all server rack IDs.

Answer

Publishing the model's marketing budget.

Answer

Sharing the home addresses of developers.

Question 10

Which TWO of the following are essential components of an 'AI Security Policy' in an enterprise?

Accepted Answer

Mandatory security and red-teaming testing for high-risk models. Option C is correct because an AI Security Policy must mandate security testing and red-teaming for high-risk models to identify vulnerabilities such as prompt injection, model evasion, data poisoning, and adversarial inputs before deployment. Option D is correct because a core purpose of the policy is to define which AI use cases are authorized and which are prohibited, giving employees clear boundaries for acceptable AI usage and data handling. Options A, B, and E are not essential components: requiring only proprietary AI models is unnecessarily restrictive and ignores secure open-source or third-party options, hardware refresh cycles belong to IT asset management rather than AI security, and disabling audit logs would undermine accountability, incident response, and forensic investigation.

Answer

Requirement to use only proprietary AI models

Answer

Definition of hardware refresh cycles for data centers

Answer

Requirement to disable all audit logs to protect privacy

Question 11

Which TWO of the following are considered 'AI Supply Chain' risks?

Accepted Answer

Inclusion of poisoned data in third-party datasets.. Option A is correct because AI supply chain risk explicitly covers the ingestion of third-party datasets that may contain poisoned or tampered training data, which can corrupt model behavior or implant backdoors. Option B is correct because pre-trained model weights downloaded from model hubs are external artifacts that can carry hidden vulnerabilities, backdoors, or malicious code, making them a core supply chain risk. Options C, D, and E are not AI supply chain risks: lack of developer training is an internal skills/awareness issue, excessive cloud storage is a resource-management concern, and high staff turnover is a human-resources/continuity issue, none of which involve compromised external components or data flowing into the AI system.

Answer

Lack of internal developer training.

Answer

Excessive use of cloud storage.

Answer

High turnover of AI research staff.

Question 12

Which TWO components are essential for a robust AI Governance framework?

Accepted Answer

Defined roles and accountability structures.. Option B is correct because a robust AI Governance framework requires clearly defined roles and accountability structures—such as an AI governance board, responsible AI officers, and data stewards—so that decision rights, oversight, and escalation paths for AI systems are unambiguous. Option E is correct because formal AI policies and standards provide the documented rules, controls, and ethical principles (for example, risk classification, model approval, bias testing, and monitoring requirements) that make governance enforceable and auditable. The remaining options do not belong: a company party schedule (A) is unrelated to governance, a list of all employee names (C) is merely a personnel roster and not a governance mechanism, and best-selling AI books (D) are general educational resources rather than essential governance components.

Answer

The company party schedule.

Answer

A list of all employee names.

Answer

The best-selling AI books.

Question 13

Which THREE practices are necessary for managing 'AI Model Version Control' as part of an overall risk management strategy?

Accepted Answer

Using Git-based tracking for model configuration and code.. Option B is correct because Git-based tracking provides immutable, auditable version history for model code, configuration files, and pipeline definitions, which is essential for reproducibility and rollback in a risk management strategy. Option C is correct because maintaining a complete lineage of training data for each model version enables traceability, bias investigation, and compliance auditing, ensuring you can reproduce and justify any deployed model. Option E is correct because automated logging of training parameters and hyperparameters (e.g., learning rate, batch size, epochs, random seeds) captures the exact conditions under which each version was produced, which is necessary for reproducibility and drift analysis. Option A is not appropriate because restricting modification to a single developer creates a single point of failure and removes peer review, weakening governance rather than strengthening version control. Option D is not appropriate because storing only the production model artifact discards prior versions, training metadata, and lineage, making rollback, auditing, and root-cause analysis impossible.

Answer

Allowing only one developer to modify the model.

Answer

Storing only the production model artifact.

Question 14

Which THREE steps are needed for 'AI Risk Assessment'?

Accepted Answer

Perform a threat modeling exercise.. An AI risk assessment must begin with discovery, so option C (Inventory the AI systems) is correct because you cannot assess risks for AI assets you have not identified, including models, datasets, and their owners. Option A (Perform a threat modeling exercise) is correct because threat modeling systematically identifies threats, attack vectors, and vulnerabilities specific to each AI system, such as data poisoning, model evasion, or prompt injection. Option D (Evaluate the risk impact and likelihood) is correct because risk assessment requires scoring each identified risk by its potential impact and probability to prioritize mitigation. Option B is not part of the assessment process; publishing results on the public internet would expose sensitive system details and is not a required step. Option E is unnecessary because training a new model is a development activity, not a validation method for an AI risk assessment.

Answer

Publish the risk assessment on the public internet.

Answer

Train a new model to verify the findings.

Question 15

Which THREE technical configurations help mitigate the risk of 'Model Exfiltration'?

Accepted Answer

Strict IAM policies limiting access to the model weight files.. Option B is correct because strict IAM policies that limit access to model weight files enforce least-privilege authorization, preventing unauthorized principals from reading or copying the stored weights that constitute the model itself. Option C is correct because API response truncation or fuzzing reduces the information returned by inference endpoints, making it harder for an attacker to reconstruct or extract internal weights through repeated querying (a model-stealing/extraction attack). Option D is correct because egress traffic filtering from the model inference environment blocks outbound connections, stopping an attacker or compromised workload from exfiltrating model artifacts or weights to external destinations. Option A is not correct because increasing the number of public endpoints expands the attack surface and makes exfiltration easier, not harder. Option E is not correct because clear-text logging of model data exposes sensitive information and increases the risk of leakage rather than mitigating exfiltration.

Answer

Increasing the number of public endpoints.

Answer

Using clear text for all model logs.