Practice Exams:

View All

Amazon AWS AI Practitioner AIF-C01 Secure AI Data Prompt Injection Privacy And Grounding Practice Test

AIF-C01 skills 5.1 | 28 original questions

This AWS Certified AI Practitioner AIF-C01 practice test focuses on secure ai data prompt injection privacy and grounding through original scenario-based questions aligned to AWS Exam Guide version 1.1 published April 30, 2026. Use the full ExamSnap AIF-C01 collection for broader practice across all five current exam domains. For broader exam preparation, review the Amazon AWS Certified AI Practitioner AIF-C01 Exam Dumps page.

Instructions: Select the best answer for each question. Review the rationale after answering. Each distractor includes a brief explanation of why it is not the strongest fit for the stated scenario.

Question 1

The risk manager at Fourth Coffee is preparing a recommendation for a contact-center transformation. The recommendation must retrieve trusted source material and include it in context so the model can answer from current evidence. Which choice is the best match? A short pilot window means the team prefers an approach that can be evaluated with clear success criteria. The project has 9 downstream consumers and a monthly review of approximately 119 sampled interactions.

RAG citation and grounding
RAG grounding
Confidence scoring
Update the knowledge source instead of fine-tuning
Human verification for high-impact claims

Correct answer: B

Why: Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy. It directly addresses the requirement in this scenario.

Option review:

A: Grounded retrieval can improve traceability and reduce unsupported generation when sources are reliable. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy. It directly addresses the requirement in this scenario.

C: Confidence signals can support routing and human review, but they must be validated for the actual application. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: RAG is a strong fit for dynamic knowledge that must be refreshed independently of the model. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Human review is appropriate when residual hallucination risk is unacceptable. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: RAG grounding – Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy.

Question 2

Margie Travel has completed discovery for a operations automation program. Before implementation, the security architect must decide how to check completeness, correctness, freshness, and consistency before data is used by the AI system. Which choice best satisfies that requirement? The architecture board will reject a choice that addresses a different problem from the one described. The rollout spans 6 application teams, each using the same approved requirement set for the next 156 days.

Audit logging
Output filtering and validation
Data-quality validation
Threat detection
Data-loss prevention

Correct answer: C

Why: Poor data quality can create both security and reliability problems. It directly addresses the requirement in this scenario.

Option review:

A: Audit trails support accountability, troubleshooting, and compliance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Validation prevents unsafe, malformed, or policy-violating outputs from directly affecting users or systems. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Poor data quality can create both security and reliability problems. It directly addresses the requirement in this scenario.

D: AI workloads require the same continuous security monitoring expected of other production systems. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: DLP controls reduce accidental or malicious data exfiltration. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Data-quality validation – Poor data quality can create both security and reliability problems.

Question 3

While planning a sales-assistant rollout, School of Fine Art identifies this requirement: treat user and retrieved content as untrusted, separate it from trusted instructions, and constrain downstream actions. Which option should the risk manager prioritize if the goal is to limit exposure of sensitive data? Budget has been approved for the project, but the team still wants to avoid unnecessary recurring consumption. The evaluation set contains examples from 3 business workflows and 193 recent production cases.

Application security
Prompt poisoning
Privacy-enhancing technology
Prompt-injection defense
Data-quality validation

Correct answer: D

Why: Prompt injection can manipulate model behavior, so applications need trust boundaries and enforceable authorization beyond prompting. It directly addresses the requirement in this scenario.

Option review:

A: Most AI systems are applications first and inherit common application-security risks. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Poisoning introduces harmful content into data or context used to shape model behavior. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Prompt injection can manipulate model behavior, so applications need trust boundaries and enforceable authorization beyond prompting. It directly addresses the requirement in this scenario.

E: Poor data quality can create both security and reliability problems. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Prompt-injection defense – Prompt injection can manipulate model behavior, so applications need trust boundaries and enforceable authorization beyond prompting.

Question 4

A proof of concept at Northwind Analytics exposed a design decision for the security architect: the solution must verify critical generated facts or structured fields against rules, databases, schemas, or external evidence before use. Which option most directly solves that problem? The team will validate the result with representative production examples before rollout. The initial rollout covers 230 internal users across 8 business units.

Retrieval Augmented Generation (RAG)
Confidence scoring
Amazon Bedrock Knowledge Bases
Update the knowledge source instead of fine-tuning
Output validation

Correct answer: E

Why: Validation catches outputs that are inconsistent with authoritative data or format requirements. It directly addresses the requirement in this scenario.

Option review:

A: RAG retrieves relevant content at request time and adds it to the model context. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Confidence signals can support routing and human review, but they must be validated for the actual application. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Knowledge Bases for Amazon Bedrock can manage ingestion, retrieval, and grounding workflows for RAG. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Validation catches outputs that are inconsistent with authoritative data or format requirements. It directly addresses the requirement in this scenario.

Learning point: Output validation – Validation catches outputs that are inconsistent with authoritative data or format requirements.

Question 5

Litware Financial is documenting the target state for a document-intelligence project. The risk manager needs a solution that can allow only authorized identities and workloads to read or modify the data required for their task. Which option is the strongest fit? The pilot has representative data, and the team will measure the selected approach against an agreed acceptance threshold. The workload processes about 267 requests during its busiest hour and has a documented fallback path.

Least-privilege data access
Infrastructure protection
Data-loss prevention
Encryption at rest and in transit
Output filtering and validation

Correct answer: A

Why: Access control limits the blast radius of compromised users, agents, or applications. It directly addresses the requirement in this scenario.

Option review:

A: Access control limits the blast radius of compromised users, agents, or applications. It directly addresses the requirement in this scenario.

B: Infrastructure controls reduce attack surface around AI components. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: DLP controls reduce accidental or malicious data exfiltration. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Encryption is a baseline data-protection control. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Validation prevents unsafe, malformed, or policy-violating outputs from directly affecting users or systems. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Least-privilege data access – Access control limits the blast radius of compromised users, agents, or applications.

Question 6

Datum Research is reviewing a knowledge-assistant rollout. The security architect has one primary requirement: record AI interactions, tool calls, administrative changes, and security-relevant events for investigation and governance. Which choice best fits the requirement? The team will document the rationale for auditors and wants the recommendation to be defensible from the scenario facts. The pilot uses 304 representative records from 2 approved data sources.
Data minimization
Audit logging
Privacy-enhancing technology
Least-privilege data access
Prompt-injection defense

Correct answer: B

Why: Audit trails support accountability, troubleshooting, and compliance. It directly addresses the requirement in this scenario.

Option review:

A: Minimization reduces privacy impact and exposure. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Audit trails support accountability, troubleshooting, and compliance. It directly addresses the requirement in this scenario.

D: Access control limits the blast radius of compromised users, agents, or applications. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Prompt injection can manipulate model behavior, so applications need trust boundaries and enforceable authorization beyond prompting. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Audit logging – Audit trails support accountability, troubleshooting, and compliance.

Question 7

During a design review for Coho Winery, the risk manager must use a calibrated confidence or support signal to decide when the system should answer, warn, or escalate. The team also wants to reduce manual handling. What should the team choose? The team wants the least complex technically correct choice that satisfies the requirement. The first release supports 7 departments and is reviewed every 341 days.

Human verification for high-impact claims
RAG citation and grounding
Confidence scoring
RAG grounding
Output validation

Correct answer: C

Why: Confidence signals can support routing and human review, but they must be validated for the actual application. It directly addresses the requirement in this scenario.

Option review:

A: Human review is appropriate when residual hallucination risk is unacceptable. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Grounded retrieval can improve traceability and reduce unsupported generation when sources are reliable. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Confidence signals can support routing and human review, but they must be validated for the actual application. It directly addresses the requirement in this scenario.

D: Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Validation catches outputs that are inconsistent with authoritative data or format requirements. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Confidence scoring – Confidence signals can support routing and human review, but they must be validated for the actual application.

Question 8

Lucerne Retail is moving a personalization program from pilot to production. The key decision is how to detect or prevent unauthorized alteration of training, retrieval, configuration, or evaluation data. Which option is the strongest fit if the team wants to use current managed AWS capabilities? The workload has passed basic feasibility checks, so the remaining question is which approach best matches the requirement. The service has a 378-millisecond internal response target for the affected workflow.

Toxicity filtering
Audit logging
Prompt-injection defense
Data-integrity control
Infrastructure protection

Correct answer: D

Why: Integrity controls reduce poisoning and tampering risk. It directly addresses the requirement in this scenario.

Option review:

A: Toxicity controls are one layer of output safety. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Audit trails support accountability, troubleshooting, and compliance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Prompt injection can manipulate model behavior, so applications need trust boundaries and enforceable authorization beyond prompting. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Integrity controls reduce poisoning and tampering risk. It directly addresses the requirement in this scenario.

E: Infrastructure controls reduce attack surface around AI components. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Data-integrity control – Integrity controls reduce poisoning and tampering risk.

Question 9

A workshop at Tailspin Toys focuses on a single decision: how to validate inputs, authenticate requests, authorize actions, manage secrets, and apply secure coding practices around model calls. Which option should the risk manager recommend? Stakeholders have ruled out a broad redesign and want the choice that most precisely addresses the stated need. The team is comparing 9 candidate designs after a 415-day proof of concept.

Infrastructure protection
Least-privilege data access
Audit logging
Prompt injection
Application security

Correct answer: E

Why: Most AI systems are applications first and inherit common application-security risks. It directly addresses the requirement in this scenario.

Option review:

A: Infrastructure controls reduce attack surface around AI components. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Access control limits the blast radius of compromised users, agents, or applications. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Audit trails support accountability, troubleshooting, and compliance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Prompt injection attacks the instruction hierarchy by embedding malicious directions in user or retrieved content. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Most AI systems are applications first and inherit common application-security risks. It directly addresses the requirement in this scenario.

Learning point: Application security – Most AI systems are applications first and inherit common application-security risks.

Question 10

For the fraud-review pilot at City Power and Light, stakeholders need to require qualified review before consequential generated content is acted upon. Which concept, service, or technique most directly addresses this goal? Operational ownership is already assigned, so the team is comparing technical fit rather than staffing models. The control owner requires evidence from 6 test groups before the 452-day release review.

Human verification for high-impact claims
Retrieval Augmented Generation (RAG)
Confidence scoring
Amazon Bedrock Knowledge Bases
RAG grounding

Correct answer: A

Why: Human review is appropriate when residual hallucination risk is unacceptable. It directly addresses the requirement in this scenario.

Option review:

A: Human review is appropriate when residual hallucination risk is unacceptable. It directly addresses the requirement in this scenario.

B: RAG retrieves relevant content at request time and adds it to the model context. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Knowledge Bases for Amazon Bedrock can manage ingestion, retrieval, and grounding workflows for RAG. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Human verification for high-impact claims – Human review is appropriate when residual hallucination risk is unacceptable.

Question 11

Consolidated Messenger is comparing alternatives for its analytics modernization. The risk manager needs to collect and send only the sensitive data that is necessary for the AI task. Which option is most appropriate while trying to meet a strict latency target? Existing application interfaces can accommodate any of the listed choices, so functional fit is the deciding factor. The project has 3 downstream consumers and a monthly review of approximately 489 sampled interactions.

Least-privilege data access
Data minimization
Privacy-enhancing technology
Vulnerability management
Application security

Correct answer: B

Why: Minimization reduces privacy impact and exposure. It directly addresses the requirement in this scenario.

Option review:

A: Access control limits the blast radius of compromised users, agents, or applications. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Minimization reduces privacy impact and exposure. It directly addresses the requirement in this scenario.

D: Model security does not replace ordinary software vulnerability management. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Most AI systems are applications first and inherit common application-security risks. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Data minimization – Minimization reduces privacy impact and exposure.

Question 12

An architecture review at Nod Publishers has narrowed a compliance-assistant prototype decision to one requirement: check generated content before it is displayed or used to trigger downstream actions. What should the security architect select? Assume the required AWS capabilities are available in the selected Region and normal governance controls are in place. The rollout spans 8 application teams, each using the same approved requirement set for the next 526 days.

Data-integrity control
Prompt-injection defense
Output filtering and validation
Privacy-enhancing technology
Threat detection

Correct answer: C

Why: Validation prevents unsafe, malformed, or policy-violating outputs from directly affecting users or systems. It directly addresses the requirement in this scenario.

Option review:

A: Integrity controls reduce poisoning and tampering risk. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Prompt injection can manipulate model behavior, so applications need trust boundaries and enforceable authorization beyond prompting. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Validation prevents unsafe, malformed, or policy-violating outputs from directly affecting users or systems. It directly addresses the requirement in this scenario.

D: Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: AI workloads require the same continuous security monitoring expected of other production systems. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Output filtering and validation – Validation prevents unsafe, malformed, or policy-violating outputs from directly affecting users or systems.

Question 13

The risk manager at Fabrikam Health is preparing a recommendation for a forecasting initiative. The recommendation must retrieve trusted source material and include it in context so the model can answer from current evidence. Which choice is the best match? The review committee wants a direct mapping from the requirement to the chosen capability. The evaluation set contains examples from 5 business workflows and 563 recent production cases.

Update the knowledge source instead of fine-tuning
RAG citation and grounding
Confidence scoring
RAG grounding
Human verification for high-impact claims

Correct answer: D

Why: Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy. It directly addresses the requirement in this scenario.

Option review:

A: RAG is a strong fit for dynamic knowledge that must be refreshed independently of the model. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy. It directly addresses the requirement in this scenario.

E: Human review is appropriate when residual hallucination risk is unacceptable. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: RAG grounding – Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy.

Question 14

Wingtip Logistics has completed discovery for a customer-support modernization. Before implementation, the security architect must decide how to reduce exposure of sensitive information using techniques appropriate to the data and use case. Which choice best satisfies that requirement? The solution will serve multiple internal teams, so the recommendation should be reusable without changing the core requirement. The initial rollout covers 600 internal users across 2 business units.

Data-integrity control
Least-privilege data access
Encryption at rest and in transit
Vulnerability management
Privacy-enhancing technology

Correct answer: E

Why: Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained. It directly addresses the requirement in this scenario.

Option review:

A: Integrity controls reduce poisoning and tampering risk. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Encryption is a baseline data-protection control. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Model security does not replace ordinary software vulnerability management. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained. It directly addresses the requirement in this scenario.

Learning point: Privacy-enhancing technology – Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained.

Question 15

While planning a agentic workflow trial, Trey Research identifies this requirement: protect model inputs, outputs, datasets, and logs from network or storage disclosure. Which option should the risk manager prioritize if the goal is to limit exposure of sensitive data? The decision must follow the workload characteristics rather than a preference for the largest model or newest service. The workload processes about 637 requests during its busiest hour and has a documented fallback path.

Encryption at rest and in transit
Privacy-enhancing technology
Audit logging
Jailbreaking
Output filtering and validation

Correct answer: A

Why: Encryption is a baseline data-protection control. It directly addresses the requirement in this scenario.

Option review:

A: Encryption is a baseline data-protection control. It directly addresses the requirement in this scenario.

B: Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Audit trails support accountability, troubleshooting, and compliance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Jailbreaks attempt to elicit behavior the system is designed to block. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Encryption at rest and in transit – Encryption is a baseline data-protection control.

Question 16

A proof of concept at Bellows College exposed a design decision for the security architect: the solution must use a calibrated confidence or support signal to decide when the system should answer, warn, or escalate. Which option most directly solves that problem? The security baseline is already defined; the decision here concerns the specific capability described in the requirement. The pilot uses 674 representative records from 4 approved data sources.

Human verification for high-impact claims
Confidence scoring
Retrieval Augmented Generation (RAG)
Update the knowledge source instead of fine-tuning
Amazon Bedrock Knowledge Bases

Correct answer: B

Why: Confidence signals can support routing and human review, but they must be validated for the actual application. It directly addresses the requirement in this scenario.

Option review:

A: Human review is appropriate when residual hallucination risk is unacceptable. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Confidence signals can support routing and human review, but they must be validated for the actual application. It directly addresses the requirement in this scenario.

C: RAG retrieves relevant content at request time and adds it to the model context. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Knowledge Bases for Amazon Bedrock can manage ingestion, retrieval, and grounding workflows for RAG. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Confidence scoring – Confidence signals can support routing and human review, but they must be validated for the actual application.

Question 17

Blue Yonder Airlines is documenting the target state for a operations automation program. The risk manager needs a solution that can check completeness, correctness, freshness, and consistency before data is used by the AI system. Which option is the strongest fit? The recommendation must solve the stated requirement without introducing unrelated platform complexity. The first release supports 9 departments and is reviewed every 711 days.

Vulnerability management
Data-loss prevention
Data-quality validation
Data minimization
Threat detection

Correct answer: C

Why: Poor data quality can create both security and reliability problems. It directly addresses the requirement in this scenario.

Option review:

A: Model security does not replace ordinary software vulnerability management. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: DLP controls reduce accidental or malicious data exfiltration. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Poor data quality can create both security and reliability problems. It directly addresses the requirement in this scenario.

D: Minimization reduces privacy impact and exposure. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Data-quality validation – Poor data quality can create both security and reliability problems.

Question 18

Woodgrove Bank is reviewing a sales-assistant rollout. The security architect has one primary requirement: detect and block sensitive information from being sent to or returned by AI workflows when policy requires. Which choice best fits the requirement? The design must remain supportable after launch, but no additional feature is required beyond the stated need. The service has a 748-millisecond internal response target for the affected workflow.

Prompt-injection defense
Least-privilege data access
Prompt hijacking
Data-loss prevention
Data-quality validation

Correct answer: D

Why: DLP controls reduce accidental or malicious data exfiltration. It directly addresses the requirement in this scenario.

Option review:

A: Prompt injection can manipulate model behavior, so applications need trust boundaries and enforceable authorization beyond prompting. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Hijacking redirects the application behavior by exploiting model instructions and context. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: DLP controls reduce accidental or malicious data exfiltration. It directly addresses the requirement in this scenario.

E: Poor data quality can create both security and reliability problems. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Data-loss prevention – DLP controls reduce accidental or malicious data exfiltration.

Question 19

During a design review for Wide World Importers, the risk manager must verify critical generated facts or structured fields against rules, databases, schemas, or external evidence before use. The team also wants to reduce manual handling. What should the team choose? A short pilot window means the team prefers an approach that can be evaluated with clear success criteria. The team is comparing 3 candidate designs after a 785-day proof of concept.

Human verification for high-impact claims
Update the knowledge source instead of fine-tuning
Confidence scoring
Amazon Bedrock Knowledge Bases
Output validation

Correct answer: E

Why: Validation catches outputs that are inconsistent with authoritative data or format requirements. It directly addresses the requirement in this scenario.

Option review:

A: Human review is appropriate when residual hallucination risk is unacceptable. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: RAG is a strong fit for dynamic knowledge that must be refreshed independently of the model. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Validation catches outputs that are inconsistent with authoritative data or format requirements. It directly addresses the requirement in this scenario.

Learning point: Output validation – Validation catches outputs that are inconsistent with authoritative data or format requirements.

Question 20

VanArsdel Media is moving a document-intelligence project from pilot to production. The key decision is how to detect or prevent unauthorized alteration of training, retrieval, configuration, or evaluation data. Which option is the strongest fit if the team wants to use current managed AWS capabilities? The architecture board will reject a choice that addresses a different problem from the one described. The control owner requires evidence from 8 test groups before the 822-day release review.

Data-integrity control
Prompt-injection defense
Output filtering and validation
Encryption at rest and in transit
Privacy-enhancing technology

Correct answer: A

Why: Integrity controls reduce poisoning and tampering risk. It directly addresses the requirement in this scenario.

Option review:

A: Integrity controls reduce poisoning and tampering risk. It directly addresses the requirement in this scenario.

C: Validation prevents unsafe, malformed, or policy-violating outputs from directly affecting users or systems. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Encryption is a baseline data-protection control. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Data-integrity control – Integrity controls reduce poisoning and tampering risk.

Question 21

A workshop at Datum Dynamics focuses on a single decision: how to identify and remediate vulnerable libraries, containers, hosts, and dependencies around the AI application. Which option should the risk manager recommend? Budget has been approved for the project, but the team still wants to avoid unnecessary recurring consumption. The project has 5 downstream consumers and a monthly review of approximately 859 sampled interactions.

Encryption at rest and in transit
Vulnerability management
Sensitive prompt exposure
Data-quality validation
Prompt-injection defense

Correct answer: B

Why: Model security does not replace ordinary software vulnerability management. It directly addresses the requirement in this scenario.

Option review:

A: Encryption is a baseline data-protection control. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Model security does not replace ordinary software vulnerability management. It directly addresses the requirement in this scenario.

C: Prompt content should not contain secrets and should be protected as application logic where appropriate. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Poor data quality can create both security and reliability problems. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Vulnerability management – Model security does not replace ordinary software vulnerability management.

Question 22

For the claims-processing redesign at Alpine Ski House, stakeholders need to require qualified review before consequential generated content is acted upon. Which concept, service, or technique most directly addresses this goal? The team will validate the result with representative production examples before rollout. The rollout spans 2 application teams, each using the same approved requirement set for the next 896 days.

RAG grounding
Update the knowledge source instead of fine-tuning
Human verification for high-impact claims
Output validation
RAG citation and grounding

Correct answer: C

Why: Human review is appropriate when residual hallucination risk is unacceptable. It directly addresses the requirement in this scenario.

Option review:

A: Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Human review is appropriate when residual hallucination risk is unacceptable. It directly addresses the requirement in this scenario.

D: Validation catches outputs that are inconsistent with authoritative data or format requirements. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Grounded retrieval can improve traceability and reduce unsupported generation when sources are reliable. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Human verification for high-impact claims – Human review is appropriate when residual hallucination risk is unacceptable.

Question 23

Humongous Insurance is comparing alternatives for its personalization program. The risk manager needs to reduce exposure of sensitive information using techniques appropriate to the data and use case. Which option is most appropriate while trying to meet a strict latency target? The pilot has representative data, and the team will measure the selected approach against an agreed acceptance threshold. The evaluation set contains examples from 7 business workflows and 933 recent production cases.

Output filtering and validation
Data-integrity control
Data-loss prevention
Privacy-enhancing technology
Audit logging

Correct answer: D

Why: Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained. It directly addresses the requirement in this scenario.

Option review:

A: Validation prevents unsafe, malformed, or policy-violating outputs from directly affecting users or systems. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Integrity controls reduce poisoning and tampering risk. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: DLP controls reduce accidental or malicious data exfiltration. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained. It directly addresses the requirement in this scenario.

E: Audit trails support accountability, troubleshooting, and compliance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Privacy-enhancing technology – Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained.

Question 24

An architecture review at Graphic Design Institute has narrowed a developer-productivity pilot decision to one requirement: identify or block abusive, hateful, or otherwise disallowed generated content. What should the security architect select? The team will document the rationale for auditors and wants the recommendation to be defensible from the scenario facts. The initial rollout covers 970 internal users across 4 business units.

Privacy-enhancing technology
Prompt poisoning
Infrastructure protection
Prompt hijacking
Toxicity filtering

Correct answer: E

Why: Toxicity controls are one layer of output safety. It directly addresses the requirement in this scenario.

Option review:

A: Privacy-enhancing techniques can minimize how much sensitive information is revealed or retained. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Infrastructure controls reduce attack surface around AI components. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Hijacking redirects the application behavior by exploiting model instructions and context. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Toxicity controls are one layer of output safety. It directly addresses the requirement in this scenario.

Learning point: Toxicity filtering – Toxicity controls are one layer of output safety.

Question 25

The risk manager at Relecloud is preparing a recommendation for a fraud-review pilot. The recommendation must retrieve trusted source material and include it in context so the model can answer from current evidence. Which choice is the best match? The team wants the least complex technically correct choice that satisfies the requirement. The workload processes about 47 requests during its busiest hour and has a documented fallback path.

RAG grounding
Amazon Bedrock Knowledge Bases
Retrieval Augmented Generation (RAG)
Human verification for high-impact claims
Update the knowledge source instead of fine-tuning

Correct answer: A

Why: Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy. It directly addresses the requirement in this scenario.

Option review:

A: Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy. It directly addresses the requirement in this scenario.

B: Knowledge Bases for Amazon Bedrock can manage ingestion, retrieval, and grounding workflows for RAG. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: RAG retrieves relevant content at request time and adds it to the model context. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Human review is appropriate when residual hallucination risk is unacceptable. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: RAG is a strong fit for dynamic knowledge that must be refreshed independently of the model. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: RAG grounding – Grounding with RAG can reduce unsupported generation when retrieval sources are relevant and trustworthy.

Question 26

Adventure Works Manufacturing has completed discovery for a analytics modernization. Before implementation, the security architect must decide how to allow only authorized identities and workloads to read or modify the data required for their task. Which choice best satisfies that requirement? The workload has passed basic feasibility checks, so the remaining question is which approach best matches the requirement. The pilot uses 84 representative records from 6 approved data sources.

Data-quality validation
Least-privilege data access
Data-integrity control
Audit logging
Vulnerability management

Correct answer: B

Why: Access control limits the blast radius of compromised users, agents, or applications. It directly addresses the requirement in this scenario.

Option review:

A: Poor data quality can create both security and reliability problems. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Access control limits the blast radius of compromised users, agents, or applications. It directly addresses the requirement in this scenario.

C: Integrity controls reduce poisoning and tampering risk. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Audit trails support accountability, troubleshooting, and compliance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Model security does not replace ordinary software vulnerability management. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Least-privilege data access – Access control limits the blast radius of compromised users, agents, or applications.

Question 27

While planning a compliance-assistant prototype, Proseware Services identifies this requirement: monitor for malicious activity around AI endpoints, identities, data, and supporting infrastructure. Which option should the risk manager prioritize if the goal is to limit exposure of sensitive data? Stakeholders have ruled out a broad redesign and want the choice that most precisely addresses the stated need. The first release supports 3 departments and is reviewed every 121 days.

Data-integrity control
Application security
Threat detection
Prompt hijacking
Encryption at rest and in transit

Correct answer: C

Why: AI workloads require the same continuous security monitoring expected of other production systems. It directly addresses the requirement in this scenario.

Option review:

A: Integrity controls reduce poisoning and tampering risk. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Most AI systems are applications first and inherit common application-security risks. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: AI workloads require the same continuous security monitoring expected of other production systems. It directly addresses the requirement in this scenario.

E: Encryption is a baseline data-protection control. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Threat detection – AI workloads require the same continuous security monitoring expected of other production systems.

Question 28

A proof of concept at Lucerne Publishing exposed a design decision for the security architect: the solution must require qualified review before consequential generated content is acted upon. Which option most directly solves that problem? Operational ownership is already assigned, so the team is comparing technical fit rather than staffing models. The service has a 158-millisecond internal response target for the affected workflow.

Confidence scoring
Output validation
Amazon Bedrock Knowledge Bases
Human verification for high-impact claims
RAG citation and grounding

Correct answer: D

Why: Human review is appropriate when residual hallucination risk is unacceptable. It directly addresses the requirement in this scenario.

Option review:

A: Confidence signals can support routing and human review, but they must be validated for the actual application. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Validation catches outputs that are inconsistent with authoritative data or format requirements. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Human review is appropriate when residual hallucination risk is unacceptable. It directly addresses the requirement in this scenario.

Learning point: Human verification for high-impact claims – Human review is appropriate when residual hallucination risk is unacceptable.

Question 1

Question 2

Question 3

Question 4

Question 5

Question 6

Question 7

Question 8

Question 9

Question 10

Question 11

Question 12

Question 13

Question 14

Question 15

Question 16

Question 17

Question 18

Question 19

Question 20

Question 21

Question 22

Question 23

Question 24

Question 25

Question 26

Question 27

Question 28

Related posts: