205 questions total · 3pages · All types, answers revealed
Page 1 of 3
Page 2Which of these is a 'Proactive' risk management strategy?
These are proactive risk management strategies.
Why this answer
Adversarial testing and threat modeling are proactive, as they identify risks before exploitation.
During an adversarial training cycle, you notice the model's accuracy on the validation set drops significantly despite high training accuracy. Which technique should you apply to mitigate potential model poisoning?
DP-SGD reduces the influence of individual training points, neutralizing poisoning attempts.
Why this answer
Differential privacy helps ensure that the model does not memorize specific training data samples, which is a common target for poisoning attacks aimed at steering model output.
What is the benefit of 'Versioning' your AI models in a production environment?
Rollback capability is a key requirement for resilience.
Why this answer
Versioning enables rapid rollback to a known good state if a new version is found to be insecure or perform poorly.
Which TWO of the following are recommended when auditing an 'AI Model' for bias and fairness?
Disaggregated metrics reveal bias.
Why this answer
Auditing for fairness requires checking data representation and measuring model performance across different demographic subgroups.
When conducting a risk assessment for an AI system, which TWO of the following are considered 'AI System' components that require individual threat assessment?
Data pipelines are critical components of AI system risk.
Why this answer
NIST and ISO standards suggest assessing both the data pipeline and the model environment (inference and training).
You are configuring private connectivity for an AI model using AWS PrivateLink. Which component is required to ensure that model inference traffic never traverses the public internet?
Interface endpoints provide private access to service APIs without public IPs.
Why this answer
An interface VPC endpoint (AWS PrivateLink) allows private connectivity to SageMaker endpoints from within a VPC.
An attacker is performing a 'Model Extraction' attack. What is the most likely goal of the attacker?
The objective is to clone the model's functionality and logic.
Why this answer
Model extraction (or theft) aims to recreate the functionality or proprietary logic of a model by querying it and training a substitute model.
When onboarding an AI startup vendor, what is the best strategy to manage 'Model Drift'?
Contractual mechanisms are essential for vendor accountability.
Why this answer
Contractual requirements for monitoring reports and incident notifications force the vendor to maintain model performance standards.
In the context of the EU AI Act, what is the 'Provider's' obligation regarding conformity assessments for high-risk AI systems?
Providers are legally mandated to demonstrate conformity for market access.
Why this answer
Providers must ensure their systems undergo a conformity assessment and bear the CE marking before market entry.
You are implementing 'Red Teaming' for a Large Language Model (LLM). What is the main objective of this exercise?
Red teaming is a simulation to discover weaknesses before deployment.
Why this answer
AI Red Teaming identifies vulnerabilities by actively attempting to exploit the model through adversarial prompts.
What is the role of 'Homomorphic Encryption' in secure AI? (Theoretical concept)
This provides the highest level of privacy for outsourced AI.
Why this answer
Homomorphic encryption allows computation on encrypted data, enabling a provider to train or run a model on data without ever decrypting it.
What is the primary role of an 'AI Ethics Board' in an organization's governance structure?
Ensures alignment of AI applications with corporate ethics.
Why this answer
Ethics boards provide guidance on moral and societal implications beyond legal compliance, ensuring alignment with organizational values.
An organization is using RAG (Retrieval-Augmented Generation). What is the primary governance concern for the retrieved documents?
Ensures that AI only accesses data the user is authorized to see.
Why this answer
Ensuring that retrieved documents are subject to appropriate access controls is critical to prevent data leakage of sensitive or unauthorized information.
Which department should own the 'AI Ethics Policy'?
Appropriate ownership for policy.
Why this answer
Compliance or Legal usually own ethics policy, in coordination with executive leadership, to ensure corporate alignment.
Which THREE actions should be included in an AI incident response plan?
Required for prevention.
Why this answer
Communication, containment, and root cause analysis are standard requirements for effective incident management.
You are managing a model's lifecycle in SageMaker. How do you ensure that only approved model versions are deployed to production?
Registry approval states prevent unauthorized deployment.
Why this answer
The SageMaker Model Registry allows for version tracking, metadata management, and approval states to control deployment.
Which of the following is a 'Control' for mitigating the risk of 'Prompt Injection'?
Sandboxing restricts the agent's ability to pivot or cause damage.
Why this answer
A 'Sandboxed Execution Environment' for agents prevents the AI from accessing the host file system or network, limiting the impact of an injection.
When establishing a risk management framework for GenAI, which THREE factors should be prioritized to satisfy the NIST AI Risk Management Framework requirements?
Continuous monitoring is a core requirement of the NIST framework.
Why this answer
NIST AI RMF prioritizes trustworthiness, safety, and reliability through continuous monitoring and governance.
Which TWO factors are critical when assessing the risk of a new Generative AI use case?
Core risk factor.
Why this answer
Business impact and data sensitivity are the foundational elements for risk assessment in AI.
Why is it important to perform 'Model Validation' before production deployment?
Testing is crucial for quality and safety assurance.
Why this answer
Model validation ensures that the model meets performance and safety requirements, minimizing the risk of unexpected behavior.
You are assessing the risk of 'Data Poisoning'. Which stage of the machine learning lifecycle is most vulnerable to this attack?
Poisoning happens during the preparation and ingestion of training data.
Why this answer
Data poisoning occurs during the training data ingestion phase, where malicious samples are injected into the training set.
Which THREE factors increase 'AI Model Risk'?
Increases risk.
Why this answer
High complexity, lack of interpretability, and critical business impact are the top drivers of AI model risk.
Which technique is best suited to defend against 'Model Extraction' where an adversary queries a public API to build a clone of your proprietary model?
Output perturbation prevents exact response matching required for extraction.
Why this answer
Adding noise (differential privacy) to the outputs or rate-limiting makes the model's decision boundaries difficult to map precisely.
Which TWO techniques are effective for detecting and mitigating 'Model Inversion' attacks where an adversary attempts to reconstruct training data from model outputs?
Prevents attackers from gathering sufficient output samples.
Why this answer
Model inversion is mitigated by reducing the information content of model outputs and limiting query volume.
Which THREE technical controls are recommended to secure an AI model's API endpoint?
Schema validation prevents malformed data injection attacks.
Why this answer
Securing AI APIs requires defense-in-depth including rate limiting, authentication, and input sanitization.
What is 'Model Drift'?
This is the definition of model drift.
Why this answer
Model drift refers to the degradation of model performance over time because the real-world data distribution changes and no longer matches the training data.
Which risk management activity happens during the 'Design' phase of an AI project?
These should be done during design (Shift Left).
Why this answer
Threat modeling and identifying ethical/privacy requirements should occur at the design stage before development begins.
Which TWO aspects of 'Transparency' are most important for AI stakeholders?
Crucial for trust and compliance.
Why this answer
Explaining the logic behind outputs and disclosing that the user is interacting with an AI are key transparency requirements.
What is the primary risk associated with 'Overfitting' in an AI model?
Generalization failure is the primary risk of overfitting.
Why this answer
Overfitting means the model memorizes the training data rather than generalizing, leading to poor performance on new, unseen data, which is a reliability risk.
You are configuring Microsoft Defender for Cloud for an Azure OpenAI deployment. To detect prompt injection attempts effectively, which specific setting must be enabled in the 'Microsoft Defender for Cloud Apps' dashboard?
AI-SPM is the designated feature set for identifying security configuration risks and anomalous activity in AI workloads.
Why this answer
Microsoft Defender for Cloud leverages 'AI Security Posture Management' to identify anomalous prompts that resemble injection patterns.
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?
FGSM is a standard evasion attack technique for testing classifier robustness.
Why this answer
The Fast Gradient Sign Method (FGSM) is a classic white-box attack technique used to test evasion vulnerability by introducing imperceptible noise.
A company is adopting the ISO/IEC 42001 AI management system standard. What is the primary purpose of the 'Context of the Organization' clause?
It establishes the framework within which AI risks are managed.
Why this answer
This clause requires the organization to identify internal and external issues, stakeholder requirements, and the scope of the AIMS.
In the context of poisoning attacks, which technique is most effective for detecting data contamination in an automated ML training pipeline?
Poisoned data often introduces statistical anomalies that outlier detection can flag.
Why this answer
Outlier detection (or anomaly detection) is the primary mechanism to identify malicious, poisoned data points that deviate from the expected distribution of the training set.
Which TWO measures are most effective in reducing the risk of 'Hallucinations' in LLM systems?
RAG anchors the model's output to specific, verifiable source documents.
Why this answer
RAG and Grounding ensure the model relies on external, verified context rather than its internal probabilistic memory.
How does 'Continuous Monitoring' differ from 'Point-in-Time Auditing'?
Core benefit of continuous monitoring.
Why this answer
Continuous monitoring provides real-time detection of drift or threats, whereas audits are snapshots in time.
Which TWO items must be documented in an 'AI Model Card'?
Understanding failure modes is essential for risk management.
Why this answer
Model Cards are meant to document intended use and limitations to ensure transparency.
What is 'AI Red Teaming'?
This is the standard definition of Red Teaming.
Why this answer
AI Red Teaming is a structured process of testing an AI system's safety and security by attempting to find vulnerabilities through adversarial attacks.
You are managing a model deployed in AWS SageMaker and observe drift in the ground truth data. Which SageMaker feature should be configured to automatically monitor and alert on this statistical deviation?
Model Monitor tracks data quality and drifts in production endpoints.
Why this answer
SageMaker Model Monitor is designed to detect data drift and concept drift by comparing real-time inference data against the baseline training statistics.
Which mechanism best ensures 'Auditability' of AI decision-making?
Ensures a verifiable trail for auditing.
Why this answer
Comprehensive logging of inputs, model parameters, and outputs is essential to recreate and audit decisions.
An organization is deploying an AI application that processes PII. The governance policy requires a formal AI impact assessment. Which activity is the most critical first step before model training begins?
Data lineage and provenance verification are mandatory under AI governance frameworks to ensure training data integrity and compliance.
Why this answer
Data classification and privacy impact assessment must occur before model training to ensure compliance with data governance policies.
Which THREE elements are essential components of an effective AI Incident Response Plan?
Essential to stop a compromised model from producing bad output.
Why this answer
An AI-specific incident response plan must address the unique nature of AI, including specialized stakeholders and specific technical triggers.
A security manager is integrating AI governance into the existing NIST AI RMF framework. Which action best ensures alignment between AI security controls and business risk appetite?
A governance structure is required to define risk appetite and ensure AI initiatives align with business security policy.
Why this answer
Mapping AI-specific controls to the organizational risk register ensures that security measures are proportionate to the business impact of AI system failure.
Which security principle is violated when an AI model is allowed to access external data sources without sufficient authorization checks on the data itself?
The AI agent must adhere to the same authorization boundaries as the user.
Why this answer
Unauthorized data access via AI agents often results from 'Broken Access Control' where the AI inherits privileges it should not possess.
Which team should lead the definition of 'AI Acceptable Use' policies?
Ensures all organizational risks are considered.
Why this answer
Cross-functional teams including legal, privacy, security, and business stakeholders ensure the policy is balanced and enforceable.
When establishing an AI steering committee, which reporting cadence is most effective for ensuring board-level oversight of residual AI risk?
Provides the appropriate balance of strategic oversight and risk management updates.
Why this answer
Quarterly reporting aligns with standard enterprise risk management cadences for high-impact emerging technologies.
You are auditing an AI system for 'Model Explainability' (XAI). Why is XAI critical from a risk management perspective?
This is the primary risk-based utility of XAI.
Why this answer
Explainability allows organizations to understand the 'why' behind model decisions, essential for regulatory compliance and debugging bias.
Which THREE security controls are essential for protecting the integrity of the AI model training supply chain?
Verifies the authenticity and source of training data.
Why this answer
Supply chain security involves securing the training data, the model architecture, and the software dependencies.
A multinational bank is deploying a proprietary LLM. Which approach provides the most robust assurance for model supply chain security?
Provides visibility into training data, libraries, and base model lineage.
Why this answer
Software Bill of Materials (SBOM) for AI components is the industry standard for tracking dependencies and vulnerabilities in the AI lifecycle.
Which TWO security controls are recommended for protecting 'Training Data'?
Encryption is standard for protecting data at rest.
Why this answer
Training data should be protected using access controls and encryption at rest to prevent unauthorized access.
What is the primary risk of 'Over-reliance' on AI outputs?
This is the core risk of automation bias.
Why this answer
Over-reliance (automation bias) leads to users trusting AI outputs without verification, which can be disastrous if the model makes errors or hallucinates.
You are configuring Microsoft Defender for Cloud to monitor your Azure OpenAI service. Which specific resource diagnostic setting must be enabled to ensure that prompt injection attempts are logged in the Log Analytics workspace?
The RequestResponse category includes the prompt and completion data required for security analysis.
Why this answer
To detect security threats like prompt injection, you must enable diagnostic logging for the Azure OpenAI service specifically for the 'RequestResponse' category, which captures the interaction payload.
In the context of 'Adversarial Machine Learning', what is a 'Blind Spot' attack?
This targets the model's decision boundary limitations.
Why this answer
A blind spot attack exploits areas in the input space where the model's confidence is high, but its performance is actually incorrect due to poor training coverage.
Which metric is most appropriate for reporting the 'AI Security Posture' to the board?
Indicates the reach and effectiveness of the governance program.
Why this answer
Percentage of models with completed risk assessments provides a clear indicator of governance coverage across the enterprise.
Which of the following describes the purpose of 'Differential Privacy' in a machine learning system?
It protects individual privacy during model learning.
Why this answer
Differential privacy ensures that an individual's data cannot be inferred from the aggregate model output by injecting statistical noise during training.
A model is vulnerable to 'Evasion Attacks' using FGSM (Fast Gradient Sign Method). Which architectural change most effectively increases robustness?
Defensive distillation makes the model less sensitive to small gradient changes.
Why this answer
FGSM uses the model's gradients to craft perturbations; defensive distillation or adversarial training helps mask these gradients.
To prevent unauthorized access to an Amazon Bedrock model, which AWS service is the primary point of control for enforcing 'Least Privilege' at the API layer?
IAM policies manage which users or services can invoke specific Bedrock models.
Why this answer
AWS IAM (Identity and Access Management) is the primary tool for defining permissions for Bedrock API operations.
When managing shadow AI, which strategy is most effective at the enterprise level?
Enables identification and policy-based control of AI service usage.
Why this answer
A 'detect and onboard' approach using API discovery tools helps identify and bring unauthorized AI into the managed governance lifecycle.
What is the purpose of 'AI Model Inventory'?
This is the definition of AI asset management.
Why this answer
An inventory tracks all models in production to ensure they are managed, monitored, and reviewed according to organizational policy.
You are managing a PyTorch model and suspect data poisoning during the training phase. Which practice is the most effective technical control to verify training data integrity in an MLOps pipeline?
Checksums verify data integrity from storage to training ingestion.
Why this answer
Data lineage and cryptographic hashing ensure that the training dataset has not been altered maliciously.
What is a 'Side-Channel Attack' against an AI model?
This is the definition of a side-channel attack.
Why this answer
Side-channel attacks exploit information leaked through indirect channels like power consumption, timing, or memory access patterns to infer sensitive model information.
Which TWO of the following are risks associated with using public pre-trained models without further security validation?
Malicious actors can embed triggers in public models.
Why this answer
Public models may contain hidden backdoors (poisoning) or have been trained on sensitive/copyrighted data that could cause legal or compliance issues.
When assessing AI third-party vendor risk, which document is most useful for understanding the vendor's data handling practices?
The DPA is the authoritative legal document regarding data handling.
Why this answer
The Data Processing Agreement (DPA) outlines exactly how data is used, stored, and protected.
Which tool is best for managing 'Secrets' (like database passwords or API keys) used by your AI training pipeline?
Vault provides centralized, secure secret management.
Why this answer
Secrets management services (AWS Secrets Manager, HashiCorp Vault) are the secure way to manage credentials.
When using a 'Managed AI API' service, what is the best practice for handling API keys?
Secrets management is the industry standard for credential protection.
Why this answer
API keys should never be hardcoded; they should be stored in secrets managers like AWS Secrets Manager or Azure Key Vault.
Which THREE technical controls effectively manage 'Third-Party AI Vendor' risks?
You should not rely solely on vendor claims.
Why this answer
Managing third-party AI risk involves technical validation, contractual oversight, and ongoing monitoring.
What does 'Model Explainability' (XAI) primarily support in governance?
Key drivers for explainability requirements.
Why this answer
XAI supports transparency and trust by allowing humans to understand the 'why' behind model outputs.
You are configuring AWS Bedrock Guardrails to prevent the model from returning PII. Which specific configuration setting should be enabled to ensure PII filtering is applied to both prompt and response?
Correct configuration for bidirectional PII filtering in Bedrock Guardrails.
Why this answer
AWS Bedrock Guardrails allow for PII entity redaction or blocking; the configuration must be set to filter both incoming prompts and outgoing model responses to ensure comprehensive protection.
Which TWO components are essential for a robust AI Governance framework?
Fundamental for governance.
Why this answer
Policies and clearly defined roles are the foundation of any governance system.
Which THREE criteria are essential for evaluating the risk of an AI vendor's 'Model Card'?
Data provenance is key to understanding bias and legality.
Why this answer
Model Cards provide transparency regarding intended use, limitations, and performance, which are critical for risk assessment.
You are managing an AI project and notice the model is performing poorly on a subset of data representing a protected class. What is the correct next step in the risk assessment process?
Bias is often rooted in data imbalances, which must be investigated.
Why this answer
Performing a root cause analysis on the data used to train that specific segment is the correct diagnostic step.
What is the primary function of an 'AI Steering Committee'?
The charter of a steering committee.
Why this answer
To set direction, prioritize initiatives, and oversee the AI risk management strategy at the enterprise level.
An AI project is failing to meet its security requirements. Which governance process should be triggered to address this?
Standard governance process for managing non-compliant or high-risk projects.
Why this answer
The AI Risk Acceptance or Remediation process ensures that management formally decides whether to fix the issue or accept the risk, documenting the decision.
Which TWO of the following are primary risks associated with 'membership inference attacks' against an AI model?
Privacy leakage is the core consequence of membership inference.
Why this answer
Membership inference attacks aim to determine if a specific data point was used in the training set, risking privacy leakage.
What is 'Model Stealing' as a risk?
Model stealing is the exfiltration of the model itself.
Why this answer
Model stealing is the attempt to gain unauthorized access to the underlying weights of a model, often through side-channel attacks or memory scraping.
Which THREE activities should be included in an AI Security Governance Program to effectively manage model supply chain risks?
Digital signatures ensure that the model has not been tampered with by third parties.
Why this answer
Managing supply chain risks requires provenance, dependency scanning, and license verification to ensure security and compliance.
Page 1 of 3
Page 2Practice AAISM by domain
Target a specific domain to shore up weak areas.
See all domains with question counts →