Amazon AWS AIP-C01 Responsible AI Practice Test

 

Topic 15 focuses on Responsible AI, Fairness, and Transparency for the AWS Certified Generative AI Developer – Professional certification and the AIP-C01 exam, using Amazon Bedrock and AWS generative AI services and architecture scenarios where relevant. For broader exam preparation, review the AWS Certified Generative AI Developer – Professional AIP-C01 Exam Dumps page. Each question includes a concise explanation of the correct answer and the technical reason the other choices are incorrect.

Question 1

The insurance automation team needs an architecture that can show users which evidence or documents support an answer. Which choice best meets that need?

1. Lambda compliance check
2. confidence and uncertainty indicators
3. Prompt Management and Prompt Flows A/B testing
4. source attribution in generated responses

Correct Answer: D

 

Correct Answer

Answer D is correct because source attribution in generated responses is designed to show users which evidence or documents support an answer. It links or cites generated claims to retrieved source material so users can inspect supporting context.

Incorrect Answers

Answer A is incorrect because Lambda compliance check is primarily used to apply a deterministic organization-specific responsible-AI rule that is not provided directly by a managed control, which is a different requirement from the one being tested.

Answer B is incorrect because confidence and uncertainty indicators is primarily used to communicate that model outputs have varying levels of certainty, which is a different requirement from the one being tested.

Answer C is incorrect because Prompt Management and Prompt Flows A/B testing is primarily used to compare alternative prompt treatments under controlled conditions, which is a different requirement from the one being tested.

 

Question 2

The procurement automation group has a design goal to communicate that model outputs have varying levels of certainty. What should the team choose?

1. human review for high-impact decisions
2. LLM-as-a-judge evaluation
3. Amazon Bedrock agent tracing
4. confidence and uncertainty indicators

Correct Answer: D

 

Correct Answer

Answer D is correct because confidence and uncertainty indicators is designed to communicate that model outputs have varying levels of certainty. It surfaces measured confidence, uncertainty, or verification status rather than presenting every generation as equally reliable.

Incorrect Answers

Answer A is incorrect because human review for high-impact decisions is primarily used to require accountable human judgment before a GenAI recommendation triggers a consequential action, which is a different requirement from the one being tested.

Answer B is incorrect because LLM-as-a-judge evaluation is primarily used to apply a model-based rubric to score outputs at scale when deterministic metrics are insufficient, which is a different requirement from the one being tested.

Answer C is incorrect because Amazon Bedrock agent tracing is primarily used to inspect agent reasoning steps, orchestration events, and tool interactions for transparency, which is a different requirement from the one being tested.

 

Question 3

The observability engineering group needs an architecture that can inspect agent reasoning steps, orchestration events, and tool interactions for transparency. Which choice best meets that need?

1. fairness evaluation
2. user-facing AI transparency notice
3. Amazon Bedrock agent tracing
4. policy-aligned Bedrock Guardrails

Correct Answer: C

 

Correct Answer

Answer C is correct because Amazon Bedrock agent tracing is designed to inspect agent reasoning steps, orchestration events, and tool interactions for transparency. It records trace information that helps developers understand how an agent arrived at an action or result.

Incorrect Answers

Answer A is incorrect because fairness evaluation is primarily used to compare model behavior across relevant groups or scenarios for systematic disparities, which is a different requirement from the one being tested.

Answer B is incorrect because user-facing AI transparency notice is primarily used to make it clear when content or recommendations are generated or assisted by AI, which is a different requirement from the one being tested.

Answer D is incorrect because policy-aligned Bedrock Guardrails is primarily used to enforce responsible-AI constraints consistently at runtime, which is a different requirement from the one being tested.

 

Question 4

The managed services operations team is prioritizing a requirement to compare model behavior across relevant groups or scenarios for systematic disparities. Which implementation is most appropriate?

1. Prompt Management and Prompt Flows A/B testing
2. fairness evaluation
3. model-card limitation documentation
4. representative bias testing

Correct Answer: B

 

Correct Answer

Answer B is correct because fairness evaluation is designed to compare model behavior across relevant groups or scenarios for systematic disparities. It uses defined fairness criteria and representative test data instead of relying on anecdotal review.

Incorrect Answers

Answer A is incorrect because Prompt Management and Prompt Flows A/B testing is primarily used to compare alternative prompt treatments under controlled conditions, which is a different requirement from the one being tested.

Answer C is incorrect because model-card limitation documentation is primarily used to record known model limitations and intended-use boundaries for reviewers and users, which is a different requirement from the one being tested.

Answer D is incorrect because representative bias testing is primarily used to evaluate responsible-AI behavior across realistic populations, prompts, and edge cases, which is a different requirement from the one being tested.

 

Question 5

The service management platform needs an architecture that can compare alternative prompt treatments under controlled conditions. Which choice best meets that need?

1. LLM-as-a-judge evaluation
2. Prompt Management and Prompt Flows A/B testing
3. source attribution in generated responses
4. Lambda compliance check

Correct Answer: B

 

Correct Answer

Answer B is correct because Prompt Management and Prompt Flows A/B testing is designed to compare alternative prompt treatments under controlled conditions. It routes representative traffic or test cases across versions so quality and fairness outcomes can be measured.

Incorrect Answers

Answer A is incorrect because LLM-as-a-judge evaluation is primarily used to apply a model-based rubric to score outputs at scale when deterministic metrics are insufficient, which is a different requirement from the one being tested.

Answer C is incorrect because source attribution in generated responses is primarily used to show users which evidence or documents support an answer, which is a different requirement from the one being tested.

Answer D is incorrect because Lambda compliance check is primarily used to apply a deterministic organization-specific responsible-AI rule that is not provided directly by a managed control, which is a different requirement from the one being tested.

 

Question 6

The digital experience platform is prioritizing a requirement to apply a model-based rubric to score outputs at scale when deterministic metrics are insufficient. Which implementation is most appropriate?

1. confidence and uncertainty indicators
2. LLM-as-a-judge evaluation
3. human review for high-impact decisions
4. policy-aligned Bedrock Guardrails

Correct Answer: B

 

Correct Answer

Answer B is correct because LLM-as-a-judge evaluation is designed to apply a model-based rubric to score outputs at scale when deterministic metrics are insufficient. It uses a carefully defined evaluator prompt and criteria to assess dimensions such as relevance, quality, or fairness.

Incorrect Answers

Answer A is incorrect because confidence and uncertainty indicators is primarily used to communicate that model outputs have varying levels of certainty, which is a different requirement from the one being tested.

Answer C is incorrect because human review for high-impact decisions is primarily used to require accountable human judgment before a GenAI recommendation triggers a consequential action, which is a different requirement from the one being tested.

Answer D is incorrect because policy-aligned Bedrock Guardrails is primarily used to enforce responsible-AI constraints consistently at runtime, which is a different requirement from the one being tested.

 

Question 7

The cloud governance office must enforce responsible-AI constraints consistently at runtime. Which approach is the strongest fit?

1. policy-aligned Bedrock Guardrails
2. user-facing AI transparency notice
3. model-card limitation documentation
4. Amazon Bedrock agent tracing

Correct Answer: A

 

Correct Answer

Answer A is correct because policy-aligned Bedrock Guardrails is designed to enforce responsible-AI constraints consistently at runtime. It translates organizational policy into managed safety and information-handling controls applied to model interactions.

Incorrect Answers

Answer B is incorrect because user-facing AI transparency notice is primarily used to make it clear when content or recommendations are generated or assisted by AI, which is a different requirement from the one being tested.

Answer C is incorrect because model-card limitation documentation is primarily used to record known model limitations and intended-use boundaries for reviewers and users, which is a different requirement from the one being tested.

Answer D is incorrect because Amazon Bedrock agent tracing is primarily used to inspect agent reasoning steps, orchestration events, and tool interactions for transparency, which is a different requirement from the one being tested.

 

Question 8

The travel booking application is prioritizing a requirement to record known model limitations and intended-use boundaries for reviewers and users. Which implementation is most appropriate?

1. Lambda compliance check
2. representative bias testing
3. model-card limitation documentation
4. fairness evaluation

Correct Answer: C

 

Correct Answer

Answer C is correct because model-card limitation documentation is designed to record known model limitations and intended-use boundaries for reviewers and users. It documents evaluation context, caveats, and constraints so deployment decisions are informed by known risks.

Incorrect Answers

Answer A is incorrect because Lambda compliance check is primarily used to apply a deterministic organization-specific responsible-AI rule that is not provided directly by a managed control, which is a different requirement from the one being tested.

Answer B is incorrect because representative bias testing is primarily used to evaluate responsible-AI behavior across realistic populations, prompts, and edge cases, which is a different requirement from the one being tested.

Answer D is incorrect because fairness evaluation is primarily used to compare model behavior across relevant groups or scenarios for systematic disparities, which is a different requirement from the one being tested.

 

Question 9

The compliance engineering group has a design goal to apply a deterministic organization-specific responsible-AI rule that is not provided directly by a managed control. What should the team choose?

1. Lambda compliance check
2. Prompt Management and Prompt Flows A/B testing
3. human review for high-impact decisions
4. source attribution in generated responses

Correct Answer: A

 

Correct Answer

Answer A is correct because Lambda compliance check is designed to apply a deterministic organization-specific responsible-AI rule that is not provided directly by a managed control. It runs custom code to validate requests, outputs, or metadata against an explicit compliance rule.

Incorrect Answers

Answer B is incorrect because Prompt Management and Prompt Flows A/B testing is primarily used to compare alternative prompt treatments under controlled conditions, which is a different requirement from the one being tested.

Answer C is incorrect because human review for high-impact decisions is primarily used to require accountable human judgment before a GenAI recommendation triggers a consequential action, which is a different requirement from the one being tested.

Answer D is incorrect because source attribution in generated responses is primarily used to show users which evidence or documents support an answer, which is a different requirement from the one being tested.

 

Question 10

The document intelligence team is prioritizing a requirement to require accountable human judgment before a GenAI recommendation triggers a consequential action. Which implementation is most appropriate?

1. user-facing AI transparency notice
2. confidence and uncertainty indicators
3. human review for high-impact decisions
4. LLM-as-a-judge evaluation

Correct Answer: C

 

Correct Answer

Answer C is correct because human review for high-impact decisions is designed to require accountable human judgment before a GenAI recommendation triggers a consequential action. It keeps sensitive decisions from being executed solely on probabilistic model output.

Incorrect Answers

Answer A is incorrect because user-facing AI transparency notice is primarily used to make it clear when content or recommendations are generated or assisted by AI, which is a different requirement from the one being tested.

Answer B is incorrect because confidence and uncertainty indicators is primarily used to communicate that model outputs have varying levels of certainty, which is a different requirement from the one being tested.

Answer D is incorrect because LLM-as-a-judge evaluation is primarily used to apply a model-based rubric to score outputs at scale when deterministic metrics are insufficient, which is a different requirement from the one being tested.

 

Question 11

The human resources automation team needs an architecture that can make it clear when content or recommendations are generated or assisted by AI. Which choice best meets that need?

1. Amazon Bedrock agent tracing
2. user-facing AI transparency notice
3. representative bias testing
4. policy-aligned Bedrock Guardrails

Correct Answer: B

 

Correct Answer

Answer B is correct because user-facing AI transparency notice is designed to make it clear when content or recommendations are generated or assisted by AI. It sets expectations about automation, limitations, and appropriate user verification.

Incorrect Answers

Answer A is incorrect because Amazon Bedrock agent tracing is primarily used to inspect agent reasoning steps, orchestration events, and tool interactions for transparency, which is a different requirement from the one being tested.

Answer C is incorrect because representative bias testing is primarily used to evaluate responsible-AI behavior across realistic populations, prompts, and edge cases, which is a different requirement from the one being tested.

Answer D is incorrect because policy-aligned Bedrock Guardrails is primarily used to enforce responsible-AI constraints consistently at runtime, which is a different requirement from the one being tested.

 

Question 12

The global infrastructure team is prioritizing a requirement to evaluate responsible-AI behavior across realistic populations, prompts, and edge cases. Which implementation is most appropriate?

1. representative bias testing
2. model-card limitation documentation
3. fairness evaluation
4. source attribution in generated responses

Correct Answer: A

 

Correct Answer

Answer A is correct because representative bias testing is designed to evaluate responsible-AI behavior across realistic populations, prompts, and edge cases. It uses balanced test cases so fairness issues are more likely to appear before production exposure.

Incorrect Answers

Answer B is incorrect because model-card limitation documentation is primarily used to record known model limitations and intended-use boundaries for reviewers and users, which is a different requirement from the one being tested.

Answer C is incorrect because fairness evaluation is primarily used to compare model behavior across relevant groups or scenarios for systematic disparities, which is a different requirement from the one being tested.

Answer D is incorrect because source attribution in generated responses is primarily used to show users which evidence or documents support an answer, which is a different requirement from the one being tested.

 

Question 13

The enterprise search program describes a component that links or cites generated claims to retrieved source material so users can inspect supporting context. Which capability is being described?

1. source attribution in generated responses
2. Prompt Management and Prompt Flows A/B testing
3. Lambda compliance check
4. confidence and uncertainty indicators

Correct Answer: A

 

Correct Answer

Answer A is correct because the description directly matches source attribution in generated responses. source attribution in generated responses links or cites generated claims to retrieved source material so users can inspect supporting context.

Incorrect Answers

Answer B is incorrect because Prompt Management and Prompt Flows A/B testing is primarily used to compare alternative prompt treatments under controlled conditions, which is a different requirement from the one being tested.

Answer C is incorrect because Lambda compliance check is primarily used to apply a deterministic organization-specific responsible-AI rule that is not provided directly by a managed control, which is a different requirement from the one being tested.

Answer D is incorrect because confidence and uncertainty indicators is primarily used to communicate that model outputs have varying levels of certainty, which is a different requirement from the one being tested.

 

Question 14

The workflow modernization program describes a component that surfaces measured confidence, uncertainty, or verification status rather than presenting every generation as equally reliable. Which capability is being described?

1. LLM-as-a-judge evaluation
2. confidence and uncertainty indicators
3. Amazon Bedrock agent tracing
4. human review for high-impact decisions

Correct Answer: B

 

Correct Answer

Answer B is correct because the description directly matches confidence and uncertainty indicators. confidence and uncertainty indicators surfaces measured confidence, uncertainty, or verification status rather than presenting every generation as equally reliable.

Incorrect Answers

Answer A is incorrect because LLM-as-a-judge evaluation is primarily used to apply a model-based rubric to score outputs at scale when deterministic metrics are insufficient, which is a different requirement from the one being tested.

Answer C is incorrect because Amazon Bedrock agent tracing is primarily used to inspect agent reasoning steps, orchestration events, and tool interactions for transparency, which is a different requirement from the one being tested.

Answer D is incorrect because human review for high-impact decisions is primarily used to require accountable human judgment before a GenAI recommendation triggers a consequential action, which is a different requirement from the one being tested.

 

Question 15

The medical research assistant describes a component that records trace information that helps developers understand how an agent arrived at an action or result. Which capability is being described?

1. fairness evaluation
2. policy-aligned Bedrock Guardrails
3. Amazon Bedrock agent tracing
4. user-facing AI transparency notice

Correct Answer: C

 

Correct Answer

Answer C is correct because the description directly matches Amazon Bedrock agent tracing. Amazon Bedrock agent tracing records trace information that helps developers understand how an agent arrived at an action or result.

Incorrect Answers

Answer A is incorrect because fairness evaluation is primarily used to compare model behavior across relevant groups or scenarios for systematic disparities, which is a different requirement from the one being tested.

Answer B is incorrect because policy-aligned Bedrock Guardrails is primarily used to enforce responsible-AI constraints consistently at runtime, which is a different requirement from the one being tested.

Answer D is incorrect because user-facing AI transparency notice is primarily used to make it clear when content or recommendations are generated or assisted by AI, which is a different requirement from the one being tested.

 

Question 16

Within the regional cloud architecture group’s architecture, which capability matches this technical description: uses defined fairness criteria and representative test data instead of relying on anecdotal review?

1. representative bias testing
2. model-card limitation documentation
3. Prompt Management and Prompt Flows A/B testing
4. fairness evaluation

Correct Answer: D

 

Correct Answer

Answer D is correct because the description directly matches fairness evaluation. fairness evaluation uses defined fairness criteria and representative test data instead of relying on anecdotal review.

Incorrect Answers

Answer A is incorrect because representative bias testing is primarily used to evaluate responsible-AI behavior across realistic populations, prompts, and edge cases, which is a different requirement from the one being tested.

Answer B is incorrect because model-card limitation documentation is primarily used to record known model limitations and intended-use boundaries for reviewers and users, which is a different requirement from the one being tested.

Answer C is incorrect because Prompt Management and Prompt Flows A/B testing is primarily used to compare alternative prompt treatments under controlled conditions, which is a different requirement from the one being tested.

 

Question 17

The retail recommendation team documents this GenAI behavior: routes representative traffic or test cases across versions so quality and fairness outcomes can be measured. Which capability matches it?

1. Prompt Management and Prompt Flows A/B testing
2. Lambda compliance check
3. LLM-as-a-judge evaluation
4. source attribution in generated responses

Correct Answer: A

 

Correct Answer

Answer A is correct because the description directly matches Prompt Management and Prompt Flows A/B testing. Prompt Management and Prompt Flows A/B testing routes representative traffic or test cases across versions so quality and fairness outcomes can be measured.

Incorrect Answers

Answer B is incorrect because Lambda compliance check is primarily used to apply a deterministic organization-specific responsible-AI rule that is not provided directly by a managed control, which is a different requirement from the one being tested.

Answer C is incorrect because LLM-as-a-judge evaluation is primarily used to apply a model-based rubric to score outputs at scale when deterministic metrics are insufficient, which is a different requirement from the one being tested.

Answer D is incorrect because source attribution in generated responses is primarily used to show users which evidence or documents support an answer, which is a different requirement from the one being tested.

 

Question 18

The telecommunications service documents this GenAI behavior: uses a carefully defined evaluator prompt and criteria to assess dimensions such as relevance, quality, or fairness. Which capability matches it?

1. human review for high-impact decisions
2. confidence and uncertainty indicators
3. policy-aligned Bedrock Guardrails
4. LLM-as-a-judge evaluation

Correct Answer: D

 

Correct Answer

Answer D is correct because the description directly matches LLM-as-a-judge evaluation. LLM-as-a-judge evaluation uses a carefully defined evaluator prompt and criteria to assess dimensions such as relevance, quality, or fairness.

Incorrect Answers

Answer A is incorrect because human review for high-impact decisions is primarily used to require accountable human judgment before a GenAI recommendation triggers a consequential action, which is a different requirement from the one being tested.

Answer B is incorrect because confidence and uncertainty indicators is primarily used to communicate that model outputs have varying levels of certainty, which is a different requirement from the one being tested.

Answer C is incorrect because policy-aligned Bedrock Guardrails is primarily used to enforce responsible-AI constraints consistently at runtime, which is a different requirement from the one being tested.

 

Question 19

Within the marketing content platform’s architecture, which capability matches this technical description: translates organizational policy into managed safety and information-handling controls applied to model interactions?

1. model-card limitation documentation
2. user-facing AI transparency notice
3. Amazon Bedrock agent tracing
4. policy-aligned Bedrock Guardrails

Correct Answer: D

 

Correct Answer

Answer D is correct because the description directly matches policy-aligned Bedrock Guardrails. policy-aligned Bedrock Guardrails translates organizational policy into managed safety and information-handling controls applied to model interactions.

Incorrect Answers

Answer A is incorrect because model-card limitation documentation is primarily used to record known model limitations and intended-use boundaries for reviewers and users, which is a different requirement from the one being tested.

Answer B is incorrect because user-facing AI transparency notice is primarily used to make it clear when content or recommendations are generated or assisted by AI, which is a different requirement from the one being tested.

Answer C is incorrect because Amazon Bedrock agent tracing is primarily used to inspect agent reasoning steps, orchestration events, and tool interactions for transparency, which is a different requirement from the one being tested.

 

Question 20

The API platform team needs to identify a capability with this behavior: documents evaluation context, caveats, and constraints so deployment decisions are informed by known risks. What is the best match?

1. Lambda compliance check
2. representative bias testing
3. fairness evaluation
4. model-card limitation documentation

Correct Answer: D

 

Correct Answer

Answer D is correct because the description directly matches model-card limitation documentation. model-card limitation documentation documents evaluation context, caveats, and constraints so deployment decisions are informed by known risks.

Incorrect Answers

Answer A is incorrect because Lambda compliance check is primarily used to apply a deterministic organization-specific responsible-AI rule that is not provided directly by a managed control, which is a different requirement from the one being tested.

Answer B is incorrect because representative bias testing is primarily used to evaluate responsible-AI behavior across realistic populations, prompts, and edge cases, which is a different requirement from the one being tested.

Answer C is incorrect because fairness evaluation is primarily used to compare model behavior across relevant groups or scenarios for systematic disparities, which is a different requirement from the one being tested.

 

Question 21

The product recommendation platform describes a component that runs custom code to validate requests, outputs, or metadata against an explicit compliance rule. Which capability is being described?

1. human review for high-impact decisions
2. Lambda compliance check
3. source attribution in generated responses
4. Prompt Management and Prompt Flows A/B testing

Correct Answer: B

 

Correct Answer

Answer B is correct because the description directly matches Lambda compliance check. Lambda compliance check runs custom code to validate requests, outputs, or metadata against an explicit compliance rule.

Incorrect Answers

Answer A is incorrect because human review for high-impact decisions is primarily used to require accountable human judgment before a GenAI recommendation triggers a consequential action, which is a different requirement from the one being tested.

Answer C is incorrect because source attribution in generated responses is primarily used to show users which evidence or documents support an answer, which is a different requirement from the one being tested.

Answer D is incorrect because Prompt Management and Prompt Flows A/B testing is primarily used to compare alternative prompt treatments under controlled conditions, which is a different requirement from the one being tested.

 

Question 22

The audit analytics group needs to identify a capability with this behavior: keeps sensitive decisions from being executed solely on probabilistic model output. What is the best match?

1. human review for high-impact decisions
2. confidence and uncertainty indicators
3. LLM-as-a-judge evaluation
4. user-facing AI transparency notice

Correct Answer: A

 

Correct Answer

Answer A is correct because the description directly matches human review for high-impact decisions. human review for high-impact decisions keeps sensitive decisions from being executed solely on probabilistic model output.

Incorrect Answers

Answer B is incorrect because confidence and uncertainty indicators is primarily used to communicate that model outputs have varying levels of certainty, which is a different requirement from the one being tested.

Answer C is incorrect because LLM-as-a-judge evaluation is primarily used to apply a model-based rubric to score outputs at scale when deterministic metrics are insufficient, which is a different requirement from the one being tested.

Answer D is incorrect because user-facing AI transparency notice is primarily used to make it clear when content or recommendations are generated or assisted by AI, which is a different requirement from the one being tested.

 

Question 23

The AI platform governance team documents this GenAI behavior: sets expectations about automation, limitations, and appropriate user verification. Which capability matches it?

1. Amazon Bedrock agent tracing
2. policy-aligned Bedrock Guardrails
3. user-facing AI transparency notice
4. representative bias testing

Correct Answer: C

 

Correct Answer

Answer C is correct because the description directly matches user-facing AI transparency notice. user-facing AI transparency notice sets expectations about automation, limitations, and appropriate user verification.

Incorrect Answers

Answer A is incorrect because Amazon Bedrock agent tracing is primarily used to inspect agent reasoning steps, orchestration events, and tool interactions for transparency, which is a different requirement from the one being tested.

Answer B is incorrect because policy-aligned Bedrock Guardrails is primarily used to enforce responsible-AI constraints consistently at runtime, which is a different requirement from the one being tested.

Answer D is incorrect because representative bias testing is primarily used to evaluate responsible-AI behavior across realistic populations, prompts, and edge cases, which is a different requirement from the one being tested.

 

Question 24

Within the transaction monitoring platform’s architecture, which capability matches this technical description: uses balanced test cases so fairness issues are more likely to appear before production exposure?

1. representative bias testing
2. fairness evaluation
3. source attribution in generated responses
4. model-card limitation documentation

Correct Answer: A

 

Correct Answer

Answer A is correct because the description directly matches representative bias testing. representative bias testing uses balanced test cases so fairness issues are more likely to appear before production exposure.

Incorrect Answers

Answer B is incorrect because fairness evaluation is primarily used to compare model behavior across relevant groups or scenarios for systematic disparities, which is a different requirement from the one being tested.

Answer C is incorrect because source attribution in generated responses is primarily used to show users which evidence or documents support an answer, which is a different requirement from the one being tested.

Answer D is incorrect because model-card limitation documentation is primarily used to record known model limitations and intended-use boundaries for reviewers and users, which is a different requirement from the one being tested.

 

Question 25

The enterprise data marketplace includes source attribution in generated responses in its architecture. What is its primary role?

1. Compare alternative prompt treatments under controlled conditions
2. Apply a deterministic organization-specific responsible-AI rule that is not provided directly by a managed control
3. Show users which evidence or documents support an answer
4. Communicate that model outputs have varying levels of certainty

Correct Answer: C

 

Correct Answer

Answer C is correct because source attribution in generated responses is specifically used to show users which evidence or documents support an answer. It links or cites generated claims to retrieved source material so users can inspect supporting context.

Incorrect Answers

Answer A is incorrect because that requirement aligns with Prompt Management and Prompt Flows A/B testing, not source attribution in generated responses.

Answer B is incorrect because that requirement aligns with Lambda compliance check, not source attribution in generated responses.

Answer D is incorrect because that requirement aligns with confidence and uncertainty indicators, not source attribution in generated responses.