205 questions total · 3pages · All types, answers revealed
Which of the following is an example of an 'Evasion' attack on a computer vision model?
Adversarial noise targets the model's inference logic.
Why this answer
Adding adversarial noise to an image (e.g., a sticker) causes the model to misclassify it, which is the definition of evasion.
Which control is most effective against 'Data Poisoning' in a collaborative learning environment?
Robust aggregation is the primary defense against poisoning.
Why this answer
By using robust statistical aggregation techniques, the aggregator can filter out malicious/outlier updates that could poison the model.
Which THREE of the following are technical controls to prevent 'Model Extraction' attacks?
Limits the volume of queries available to reconstruct the model.
Why this answer
Model extraction is mitigated by rate limiting, output perturbation (noise), and limiting the detail/granularity of the output returned to the API.
During an internal audit of the AI program, it is found that the organization lacks a 'Model Inventory'. What is the most immediate risk to the AI security program?
A model inventory is the prerequisite for all governance, risk assessment, and lifecycle management activities.
Why this answer
Without a model inventory, security teams cannot identify which models exist, making it impossible to manage risk or apply security patches.
What is the purpose of 'AI Model Transparency' in a risk management framework?
Transparency is about informed oversight and risk assessment.
Why this answer
Transparency allows stakeholders to understand how a model functions, its limitations, and its provenance to ensure accountability.
Which TWO areas should be covered by an AI procurement policy?
Essential for legal and privacy compliance.
Why this answer
Vendor security assessment and contractual data usage clauses are critical for third-party AI risk.
You detect an unusual spike in API calls that are clearly non-human. What is the most effective way to address this without completely blocking the service?
This differentiates bots from legitimate users.
Why this answer
Implementing a CAPTCHA or sophisticated WAF challenge for high-volume users is a standard way to differentiate human traffic from bots.
Which THREE elements are essential for 'AI Governance'?
Policies are the foundation of governance.
Why this answer
AI governance requires clear policies, oversight committees, and defined risk frameworks.
Which scenario best illustrates an 'AI Security Incident'?
Compromise of model integrity/confidentiality is a security incident.
Why this answer
An unauthorized model modification or data extraction is a clear security event requiring response.
An AI security team needs to ensure that internal AI tools comply with the organization's 'Human-in-the-Loop' policy. Which architectural control best enforces this governance requirement?
This technical enforcement of a governance policy ensures that AI does not act unilaterally on sensitive business processes.
Why this answer
Implementing an approval gate in the workflow ensures that critical decisions made by AI are reviewed by humans, satisfying the governance policy.
You are managing risks for a federated learning deployment. What is the primary security concern regarding the 'model updates' sent from the edge clients to the aggregator?
Gradient leakage is the primary threat in federated learning architectures.
Why this answer
In federated learning, local model updates can be analyzed to reconstruct private training data from the clients.
When assessing a third-party AI provider, which document is most essential for due diligence?
Provides verifiable evidence of security controls.
Why this answer
The System and Organization Controls (SOC) 2 Type II report provides independent assurance of the vendor's security and privacy controls.
Which control is most effective against 'Prompt Injection' attacks?
Primary defense mechanism.
Why this answer
Implementing input sanitization and output filtering via an AI Gateway is a standard defensive control.
What is the primary objective of a 'Bias Audit' in an AI risk management program?
Detecting disparate impact is the primary goal of a bias audit.
Why this answer
Bias audits ensure that AI models do not produce discriminatory outcomes based on protected classes.
You are assessing a supply chain risk where an AI model uses a pre-trained base model from a third-party hub. Which action most effectively mitigates the risk of model poisoning during the integration phase?
Cryptographic verification ensures the model has not been altered since publication.
Why this answer
Verifying the cryptographic hash of the model weights against the provider's manifest ensures integrity before deployment.
When utilizing a third-party AI vendor, which contractual requirement is most critical for addressing 'shadow AI' risks within an enterprise?
Understanding the supply chain is vital for managing third-party AI risk and shadow deployment.
Why this answer
Standardized procurement processes and inventory tracking (shadow AI) require clear vendor auditing and accountability clauses.
Which TWO reporting formats are effective for executive AI oversight?
Effective for executive review.
Why this answer
Risk heatmaps and executive summaries provide the high-level view required by management.
Which TWO factors help maintain 'AI Accountability'?
Accountability starts with ownership.
Why this answer
Accountability is maintained through clear roles/responsibilities and robust documentation of decision-making processes.
An enterprise is defining its 'AI System Inventory'. Which metadata attribute is most critical for assessing legal and regulatory risk?
Determines the legal basis for data processing and potential copyright infringement.
Why this answer
Data origin and provenance are critical for determining compliance with GDPR, AI Act, and intellectual property requirements.
When utilizing the 'Adversarial Robustness Toolbox' (ART) to defend against evasion attacks, which attack type is specifically mitigated by using 'Adversarial Training'?
Adversarial training directly hardens models against adversarial examples crafted for evasion.
Why this answer
Adversarial training involves augmenting the training set with adversarial examples, hardening the model against evasion (inference-time) attacks.
An organization is evaluating 'Third-Party AI Models'. Which risk is most effectively addressed by a 'Model Evaluation Report'?
Evaluation reports provide the data needed to assess suitability.
Why this answer
An evaluation report details the model's capabilities, limitations, and performance, helping the user determine if it fits their risk profile.
When drafting an AI policy, how should the organization address 'Human-in-the-Loop' (HITL) requirements?
Risk-based approach allows for automation of low-risk tasks while safeguarding high-stakes decisions.
Why this answer
Policy must clearly define decision-making thresholds where automated systems require human intervention to mitigate high-impact risks.
Which of the following is a 'Human-in-the-Loop' (HITL) control used for AI risk management?
This provides a safety check for critical decisions.
Why this answer
HITL involves human review of model decisions before they are finalized or acted upon, especially for high-stakes decisions.
What is the primary role of 'Model Versioning' in AI governance?
Crucial for stability and incident management.
Why this answer
Versioning ensures that changes to the model can be tracked, reverted, and audited, which is vital for reproducibility and risk control.
An organization is using an LLM-based agent. What is the primary risk associated with 'indirect prompt injection' in this environment?
This is the definition of indirect prompt injection where the LLM executes malicious input.
Why this answer
Indirect prompt injection occurs when malicious data is embedded in a retrieved document that instructs the LLM to perform unintended actions.
A board of directors requests a summary of AI risk exposure. Which metric provides the most relevant insight into AI security program maturity?
This metric directly demonstrates the effectiveness of the governance program in controlling risk.
Why this answer
Reporting on the percentage of AI models subjected to formal security and bias audits provides a clear picture of governance coverage.
You are auditing an AI system for compliance with the EU AI Act. Which control is mandatory for 'High-Risk' AI systems to ensure transparency for human oversight?
Mandatory requirement for high-risk AI to allow human intervention and auditing.
Why this answer
The EU AI Act requires that high-risk systems be designed with human-in-the-loop (HITL) capabilities and adequate logging to ensure transparency and accountability.
Your organization uses Azure OpenAI. You need to ensure that no sensitive data (e.g., credit card numbers) is sent to the model. Which specific service feature should be enabled to intercept and mask this data?
Content Safety provides real-time PII detection and masking for inputs.
Why this answer
Azure OpenAI Content Safety provides text filtering and PII masking.
What is the primary role of 'Data Encryption at Rest' for AI datasets?
Encryption is the primary control for storage security.
Why this answer
Encryption at rest ensures that even if the storage media is stolen or accessed unauthorized, the data remains unreadable.
Which TWO activities ensure 'AI Safety' in development?
Identifies safety failures.
Why this answer
Adversarial testing and rigorous documentation are key safety practices.
When evaluating AI security, 'Model Drift' refers to performance degradation over time. Which tool would a practitioner use to detect drift in a production environment?
SageMaker Model Monitor identifies drift by comparing live data to training data.
Why this answer
Model monitoring tools track statistical properties of features and predictions to identify drift.
Which document outlines the 'Roles and Responsibilities' for AI security?
The authoritative document for governance roles.
Why this answer
The AI Governance Charter or Policy defines the organizational structure for AI risk management.
Which document should define the organization's stance on AI transparency and explainability?
Policy sets the standards for transparency and compliance.
Why this answer
The AI Policy is the foundational document that sets the requirements for transparency, accountability, and explainability across the enterprise.
You are configuring a 'Firewall' for an AI API. Which rule should you prioritize to prevent unauthorized exfiltration of model training data?
Limiting outbound traffic is essential to stop exfiltration.
Why this answer
Egress filtering (outbound rules) ensures that the model environment cannot communicate with unauthorized external servers, preventing data exfiltration.
Which TWO of the following are primary security controls for preventing 'Model Inversion' attacks where an adversary attempts to reconstruct training data from model outputs?
Differential privacy mathematically bounds the influence of individual data points.
Why this answer
Differential privacy and output perturbation are the standard methods to reduce the probability of sensitive data leakage through inference queries.
An organization is updating its AI governance policy for high-risk systems. Which TWO requirements should be mandated to ensure accountability and forensic capability?
This is critical for investigating malicious use or bias in output.
Why this answer
Accountability and forensics require audit trails for both model inputs/outputs and human interactions with the system.
Which THREE factors are critical when establishing an AI risk reporting structure for executive leadership?
Reporting must connect technical AI risk to business strategy and tolerance.
Why this answer
Executive reporting must focus on business impact, risk posture, and compliance with the regulatory landscape.
Which phase of the AI Lifecycle is most critical for 'Bias Detection'?
Critical phase for bias mitigation and validation.
Why this answer
Testing and evaluation (pre-deployment) is the primary window for detecting and mitigating bias before it impacts users.
Which THREE items should be included in an 'AI Inventory'?
Essential inventory metadata.
Why this answer
Model version, owner, and risk rating are the core pieces of information needed to manage an AI portfolio.
An AI security lead is configuring a 'Model Risk Management' (MRM) framework. What is the main objective of a 'Model Validation' report?
Validation acts as a second line of defense to ensure model reliability.
Why this answer
Validation provides independent verification that the model performs as expected within its defined scope and limitations.
An organization is using Google Cloud Vertex AI and suspects an adversarial poisoning attack on a custom training dataset. Which tool should be used to verify the integrity and provenance of the training artifacts?
Provides tracking for artifacts and lineage, essential for identifying injected malicious training data.
Why this answer
Vertex AI Model Garden and Metadata are used for lineage tracking, but specifically for provenance and audit, the Vertex AI ML Metadata service allows tracking of dataset versions and training pipeline execution history.
When conducting an AI audit, what should be the primary focus of the 'Data Governance' review?
These are fundamental components of data governance.
Why this answer
Data governance ensures that only authorized, high-quality, and compliant data is used for training and inference.
When an AI model is suspected of being involved in a security incident, what is the first step in the incident response process?
Containment is standard practice for incident response.
Why this answer
Isolation is the first step to prevent further damage or exfiltration of sensitive data.
When evaluating an AI model for production, what does the 'F1 Score' tell you from a risk management perspective?
F1 is the harmonic mean of precision and recall, essential for risk evaluation.
Why this answer
The F1 score provides a balanced view of precision and recall, helping assess the risk of both false positives and false negatives.
When implementing Google Cloud's 'Vertex AI' model guardrails, you need to prevent the leakage of PII in model outputs. Which configuration tool is used to enforce these constraints at the request/response level?
These filters allow developers to set threshold levels for various categories of sensitive content.
Why this answer
The Vertex AI 'Safety Settings' and 'Content filtering' allow practitioners to define thresholds for harassment, hate speech, and PII leakage.
A machine learning model is showing signs of 'data drift'. What is the most effective initial step in the risk assessment process?
Statistical analysis is the baseline step for confirming and scoping data drift.
Why this answer
Data drift monitoring involves comparing current production data distributions against training data distributions to detect performance decay.
Which TWO actions should be taken to enforce Least Privilege access for developers interacting with SageMaker Model Monitor?
Limiting this action prevents unauthorized creation of monitoring jobs.
Why this answer
Access to Model Monitor requires permissions to both the infrastructure (S3) and the specific model monitoring job definitions.
Which THREE factors must be included in a 'Model Risk Management' (MRM) policy?
Governance gates are essential for ensuring risks are reviewed.
Why this answer
A robust MRM policy covers the lifecycle from model development and validation to monitoring and retirement.
What is the key risk difference between 'Data Privacy' and 'Model Security' in AI?
This is the correct distinction between the two domains.
Why this answer
Data privacy focuses on the protection of the information used in models (PII), while model security focuses on the integrity/availability of the model logic.
In the context of AI governance, what is the primary role of a 'Model Card'?
Model Cards are essential for ethical AI and transparency.
Why this answer
A Model Card is a documentation artifact that provides transparency into the model's intended use, limitations, and training data.
What is the purpose of 'Model Signing' in an MLOps pipeline?
Signing proves the model's origin and that it remains unaltered.
Why this answer
Model signing uses digital signatures to verify that the model artifact has not been tampered with between the training environment and the production environment.
What is the best way to manage 'Bias' in an AI system?
This is the proactive, standard practice for managing bias.
Why this answer
Regular auditing and dataset balancing are the most effective ways to identify and mitigate bias in AI systems.
What is the primary role of a 'Human-in-the-loop' (HITL) in AI risk management?
HITL serves as a governance control for critical AI decisions.
Why this answer
HITL ensures that human oversight is present in high-stakes decisions, mitigating risks of automation bias or AI failure.
During an adversarial threat modeling exercise for a facial recognition system, you identify a risk of pixel perturbation attacks. Which countermeasure is the most appropriate technical control to implement?
Adversarial training is the standard method for improving robustness against evasion attacks.
Why this answer
Adversarial training involves training the model on examples containing perturbations, making it robust against those specific attacks.
Gateways provide central control over AI interactions.
Why this answer
An LLM gateway acts as a proxy to enforce security policies, redact PII, and prevent prompt injection before requests reach the model.
An AI administrator is using Azure AI Content Safety to detect jailbreak attempts in prompt engineering. Which configuration must be enabled to proactively block prompt injection attacks?
Prompt Shield is the designated control for blocking jailbreak and prompt injection attempts.
Why this answer
The Prompt Shield feature in Azure AI Content Safety specifically targets user-to-model prompt injection and jailbreak attacks.
Which TWO aspects of 'Compliance' must be addressed by an AI governance program?
Mandatory for compliance.
Why this answer
Regulatory alignment and contractual commitments are the two pillars of AI compliance.
An organization is performing threat modeling for a RAG (Retrieval-Augmented Generation) pipeline. Which specific vulnerability is unique to the retrieval component?
Retrieval poisoning is the specific risk where the context provider is compromised.
Why this answer
In RAG, if the search or retrieval index is manipulated, the LLM will generate answers based on attacker-supplied context.
What is the benefit of 'AI Model Watermarking' for governance?
Helps manage risks related to synthetic media.
Why this answer
Watermarking assists in identifying synthetic content, which is crucial for provenance and combating misinformation.
When assessing the risk of 'Prompt Leaking', what is the most important control to implement?
Filtering the output for system instructions prevents the leak.
Why this answer
Prompt leaking occurs when a model reveals its system instructions. Limiting the LLM's access to sensitive context and using robust system prompts is key.
In an AI supply chain, what is 'Dependency Confusion' risk?
This is the classic dependency confusion attack.
Why this answer
Dependency confusion occurs when an attacker publishes a malicious package with the same name as an internal private package, hoping the package manager will pull the malicious public version.
A company is performing an 'AI Impact Assessment' (AIA). Which domain is the most critical for identifying 'Privacy' risks?
These elements are central to privacy risk.
Why this answer
Data processing and consent management are the pillars of AI privacy impact assessments.
Which TWO techniques help mitigate 'Model Poisoning'?
Filtering out anomalies prevents malicious data from entering the training set.
Why this answer
Defensive techniques like data sanitization and robust aggregation algorithms help mitigate poisoning attempts.
An organization uses IBM watsonx.ai to manage model governance. Which tool is primarily used to track 'FactSheets' for model lineage and auditability?
FactSheets in watsonx.ai are the standard tool for recording model metadata and lineage.
Why this answer
The IBM watsonx.ai Governance module uses FactSheets to automatically capture and document the lifecycle, lineage, and validation status of models.
You are reviewing logs for an AI service. You notice a high frequency of '429 Too Many Requests' errors. Which security control is likely being triggered to protect the model from potential automated extraction or DoS?
Rate limiting restricts the volume of requests a client can make.
Why this answer
Rate limiting is a standard control to prevent resource exhaustion and model scraping.
You are auditing access controls for a Microsoft Azure OpenAI deployment. Which TWO of the following configurations are required to ensure identity-based access and logging for model inference?
Required for centralized logging and security auditing.
Why this answer
Azure OpenAI requires Entra ID (Azure AD) for authentication and Azure Monitor/Log Analytics for auditing request/response patterns.
Which THREE components are critical for an 'AI Incident Response' plan?
Cooperation between technical and legal is required for AI incidents.
Why this answer
An AI IR plan must account for model-specific issues like unexpected performance drops, bias, and adversarial attacks.
An AI security manager is implementing an AI governance framework using the NIST AI RMF. Which action best demonstrates the 'Govern' function for a new generative AI deployment?
Establishing clear accountability is a core component of the Govern function.
Why this answer
The 'Govern' function focuses on establishing a culture of risk management, accountability, and organizational processes to manage AI risks.
What is the primary goal of 'Model Monitoring' in the context of risk management?
This is the main risk-related reason for monitoring.
Why this answer
Continuous monitoring detects anomalies, performance degradation, and drift, which are critical triggers for risk interventions.
What is 'Model Stealing' and why is it considered a security incident?
Stealing a model's decision function undermines its competitive value.
Why this answer
Model stealing involves querying a model to extract enough information to replicate its functionality, leading to the loss of proprietary intellectual property.
An AI security program manager is assessing AI risks. Which tool provides the most centralized oversight for AI model compliance?
Provides audit trails, policy mapping, and risk tracking.
Why this answer
A GRC (Governance, Risk, and Compliance) platform configured for AI enables systematic tracking of risks and policy adherence.
Which THREE practices are necessary for managing 'AI Model Version Control' as part of an overall risk management strategy?
Git is standard for tracking configuration and infrastructure as code.
Why this answer
Version control ensures reproducibility, rollback capability, and auditability in AI risk management.
When a model fails in production, what is the first step in the 'Incident Response' process?
Standard IR procedure.
Why this answer
Identifying and containing the impact is the immediate priority to prevent further risk.
Which TWO of the following are effective methods for monitoring and detecting AI security incidents in a cloud environment?
SIEM integration allows for correlation of AI events with other security signals.
Why this answer
Centralized logging (e.g., CloudWatch, Azure Monitor) and anomaly detection are critical for identifying unexpected model interactions.
When training an AI model, why is it critical to ensure 'Data Provenance'?
Provenance is foundational to data supply chain security.
Why this answer
Data provenance allows auditors to trace the origin of data, ensuring no poisoned or copyrighted data entered the training pipeline.
Practice AAISM by domain
Target a specific domain to shore up weak areas.
See all domains with question counts →