Topic 10 Practice Test 1 covers Data Security, Privacy, Network Isolation, PII Protection, Masking, and Retention in Domain 3 (Task 3.2). For broader exam preparation, review the AIP-C01 Exam Dumps. Every option includes focused technical reasoning tied to the tested decision.
Question 1
A GenAI application in private subnets must invoke supported AWS services without sending service traffic through an internet gateway. Which approach should the developer use? Choose ONE.
- Security groups for endpoint access
- IAM resource-scoped policy
- TLS for data in transit
- VPC endpoint for private service access
Correct Answer: D
Correct Answer
Answer D is correct because VPC endpoint for private service access — in this scenario, keeps supported AWS service traffic, in this case, on private connectivity without requiring, in this case, public internet traversal. Useful when a protected GenAI environment must, at this layer, avoid public network paths. Need: keep AWS service traffic on private network paths. VPC endpoint for private service access matches it directly.
Incorrect Answers
Answer A is incorrect because Security groups for endpoint access — for this design, restrict network flows to the, for this workload, intended sources and destinations around, for this workload, VPC resources or endpoints. Fits when private connectivity exists but network-level, for the stated need, reachability still needs narrowing. Constraint: keep AWS service traffic on private network paths. Security groups for endpoint access targets another layer.
Answer B is incorrect because IAM resource-scoped policy — in this scenario, grants only the required actions, in this case, against the specific model, data,, in this case, or application resources that need, in this case, access. Useful when the requirement is authorization minimization, at this layer, rather than network isolation. Need: keep AWS service traffic on private network paths. IAM resource-scoped policy solves another problem.
Answer C is incorrect because TLS for data in transit — in this scenario, protects data while it moves, in this case, between clients and services using, in this case, encrypted transport. Useful when the threat concerns interception of, at this layer, traffic in transit. Need: keep AWS service traffic on private network paths. TLS for data in transit solves another problem.
Question 2
The application role can invoke any Bedrock model in the account, but policy allows only one approved model and one inference profile. What should the team implement? Choose ONE.
- Lake Formation fine-grained access
- VPC endpoint for private service access
- Tool allowlist
- IAM resource-scoped policy
Correct Answer: D
Correct Answer
Answer D is correct because IAM resource-scoped policy — for this design, grants only the required actions, for this workload, against the specific model, data,, for this workload, or application resources that need, for this workload, access. Fits when the requirement is authorization minimization, for the stated need, rather than network isolation. Constraint: restrict model invocation authorization to approved resources. IAM resource-scoped policy is the closest fit.
Incorrect Answers
Answer A is incorrect because Lake Formation fine-grained access — in this scenario, controls governed access to data-lake, in this case, resources at granular database, table,, in this case, column, or related policy boundaries. Useful when different GenAI consumers need controlled, at this layer, access to shared lake data. Need: restrict model invocation authorization to approved resources. Lake Formation fine-grained access solves another problem.
Answer B is incorrect because VPC endpoint for private service access — for this design, keeps supported AWS service traffic, for this workload, on private connectivity without requiring, for this workload, public internet traversal. Fits when a protected GenAI environment must, for the stated need, avoid public network paths. Constraint: restrict model invocation authorization to approved resources. VPC endpoint for private service access targets another layer.
Answer C is incorrect because Tool allowlist — under these constraints, restricts the agent to an, at this layer, approved set of callable operations, at this layer, instead of exposing every available, at this layer, action. Makes sense when the risk boundary is which, under these constraints, tools the model may invoke. The case depends on restrict model invocation authorization to approved resources. Tool allowlist handles another concern.
Question 3
A RAG application reads a governed data lake where one team may use only selected tables and columns. Which option most directly addresses the requirement? Choose ONE.
- Lake Formation fine-grained access
- IAM resource-scoped policy
- Data masking before model invocation
- S3 Block Public Access
Correct Answer: A
Correct Answer
Answer A is correct because Lake Formation fine-grained access — for this design, controls governed access to data-lake, for this workload, resources at granular database, table,, for this workload, column, or related policy boundaries. Fits when different GenAI consumers need controlled, for the stated need, access to shared lake data. Constraint: apply fine-grained governed access to shared lake data. Lake Formation fine-grained access is the closest fit.
Incorrect Answers
Answer B is incorrect because IAM resource-scoped policy — at this decision point, grants only the required actions, within this design, against the specific model, data,, within this design, or application resources that need, within this design, access. Typical use: the requirement is authorization minimization, in the described flow, rather than network isolation. Goal: apply fine-grained governed access to shared lake data. IAM resource-scoped policy does not provide it.
Answer C is incorrect because Data masking before model invocation — in this scenario, replaces sensitive values with masked, in this case, representations before they are sent, in this case, to the model. Useful when the model does not need, at this layer, the original sensitive value to, at this layer, perform the task. Need: apply fine-grained governed access to shared lake data. Data masking before model invocation solves another problem.
Answer D is incorrect because S3 Block Public Access — in this scenario, prevents public exposure paths for, in this case, S3 buckets and objects at, in this case, the account or bucket level. Useful when the key risk is accidental, at this layer, public access to S3 data. Need: apply fine-grained governed access to shared lake data. S3 Block Public Access solves another problem.
Question 4
Security operations needs alarms and telemetry when access patterns to the GenAI data layer suddenly change. Which design best satisfies this requirement? Choose ONE.
- CloudWatch access monitoring
- Amazon Macie sensitive-data discovery
- S3 Lifecycle retention policy
- AWS KMS encryption
Correct Answer: A
Correct Answer
Answer A is correct because CloudWatch access monitoring — in this scenario, collects operational telemetry and alarms, in this case, so suspicious or unexpected access, in this case, behavior can be detected. Useful when security teams need monitoring rather, at this layer, than a preventive data transformation. Need: monitor access behavior rather than transform the data. CloudWatch access monitoring matches it directly.
Incorrect Answers
Answer B is incorrect because Amazon Macie sensitive-data discovery — in this scenario, discovers and classifies sensitive data, in this case, such as PII in Amazon, in this case, S3 using managed data-security capabilities. Useful when the organization must locate sensitive, at this layer, information stored in S3 at, at this layer, scale. Need: monitor access behavior rather than transform the data. Amazon Macie sensitive-data discovery solves another problem.
Answer C is incorrect because S3 Lifecycle retention policy — in this scenario, transitions or expires S3 objects, in this case, according to age-based lifecycle rules. Useful when stored prompts, outputs, or evidence, at this layer, must be deleted or tiered, at this layer, after a defined retention period. Need: monitor access behavior rather than transform the data. S3 Lifecycle retention policy solves another problem.
Answer D is incorrect because AWS KMS encryption — in this scenario, uses customer-managed or AWS managed, in this case, cryptographic keys to protect supported, in this case, data at rest and control, in this case, key usage. Useful when the primary control is encryption, at this layer, and key governance for stored, at this layer, data. Need: monitor access behavior rather than transform the data. AWS KMS encryption solves another problem.
Question 5
Millions of S3 objects may contain names, addresses, and identification numbers; the team must discover where sensitive data resides. What is the most appropriate implementation choice? Choose ONE.
- PII classification before indexing
- Amazon Macie sensitive-data discovery
- Bedrock Guardrails sensitive-information filter
- Amazon Comprehend PII detection
Correct Answer: B
Correct Answer
Answer B is correct because Amazon Macie sensitive-data discovery — for this design, discovers and classifies sensitive data, for this workload, such as PII in Amazon, for this workload, S3 using managed data-security capabilities. Fits when the organization must locate sensitive, for the stated need, information stored in S3 at, for the stated need, scale. Constraint: discover sensitive information across S3 at scale. Amazon Macie sensitive-data discovery is the closest fit.
Incorrect Answers
Answer A is incorrect because PII classification before indexing — in this scenario, detects and labels sensitive content, in this case, before it enters a vector, in this case, or retrieval index. Useful when retrieval should not expose sensitive, at this layer, chunks that should never have, at this layer, been indexed. Need: discover sensitive information across S3 at scale. PII classification before indexing solves another problem.
Answer C is incorrect because Bedrock Guardrails sensitive-information filter — in this scenario, can detect and block or, in this case, mask configured sensitive information in, in this case, model interactions. Useful when sensitive data must be filtered, at this layer, at the GenAI interaction boundary. Need: discover sensitive information across S3 at scale. Bedrock Guardrails sensitive-information filter solves another problem.
Answer D is incorrect because Amazon Comprehend PII detection — in this scenario, detects PII entities in text, in this case, so an application can identify, in this case, sensitive text before or during, in this case, processing. Useful when the application must find PII, at this layer, in text content programmatically. Need: discover sensitive information across S3 at scale. Amazon Comprehend PII detection solves another problem.
Question 6
A chat application must find person names, addresses, and other PII in free-form text before deciding whether to redact it. Which approach should the developer use? Choose ONE.
- Bedrock Guardrails sensitive-information filter
- Amazon Comprehend PII detection
- Amazon Macie sensitive-data discovery
- Data minimization
Correct Answer: B
Correct Answer
Answer B is correct because Amazon Comprehend PII detection — for this design, detects PII entities in text, for this workload, so an application can identify, for this workload, sensitive text before or during, for this workload, processing. Fits when the application must find PII, for the stated need, in text content programmatically. Constraint: detect PII entities programmatically in text. Amazon Comprehend PII detection is the closest fit.
Incorrect Answers
Answer A is incorrect because Bedrock Guardrails sensitive-information filter — for this design, can detect and block or, for this workload, mask configured sensitive information in, for this workload, model interactions. Fits when sensitive data must be filtered, for the stated need, at the GenAI interaction boundary. Constraint: detect PII entities programmatically in text. Bedrock Guardrails sensitive-information filter targets another layer.
Answer C is incorrect because Amazon Macie sensitive-data discovery — at this decision point, discovers and classifies sensitive data, within this design, such as PII in Amazon, within this design, S3 using managed data-security capabilities. Typical use: the organization must locate sensitive, in the described flow, information stored in S3 at, in the described flow, scale. Goal: detect PII entities programmatically in text. Amazon Macie sensitive-data discovery does not provide it.
Answer D is incorrect because Data minimization — in this scenario, limits collection and prompt context, in this case, to only the data required, in this case, for the stated purpose. Useful when privacy risk should be reduced, at this layer, before applying downstream controls. Need: detect PII entities programmatically in text. Data minimization solves another problem.
Question 7
Model responses must mask configured sensitive-information patterns before they are returned to users. What should the team implement? Choose ONE.
- Amazon Comprehend PII detection
- AWS KMS encryption
- Bedrock Guardrails sensitive-information filter
- Post-output PII filter
Correct Answer: C
Correct Answer
Answer C is correct because Bedrock Guardrails sensitive-information filter — at this decision point, can detect and block or, within this design, mask configured sensitive information in, within this design, model interactions. Typical use: sensitive data must be filtered, in the described flow, at the GenAI interaction boundary. Goal: apply a GenAI-boundary filter to sensitive output. Bedrock Guardrails sensitive-information filter provides that behavior.
Incorrect Answers
Answer A is incorrect because Amazon Comprehend PII detection — at this decision point, detects PII entities in text, within this design, so an application can identify, within this design, sensitive text before or during, within this design, processing. Typical use: the application must find PII, in the described flow, in text content programmatically. Goal: apply a GenAI-boundary filter to sensitive output. Amazon Comprehend PII detection does not provide it.
Answer B is incorrect because AWS KMS encryption — for this design, uses customer-managed or AWS managed, for this workload, cryptographic keys to protect supported, for this workload, data at rest and control, for this workload, key usage. Fits when the primary control is encryption, for the stated need, and key governance for stored, for the stated need, data. Constraint: apply a GenAI-boundary filter to sensitive output. AWS KMS encryption targets another layer.
Answer D is incorrect because Post-output PII filter — in this scenario, checks generated content for sensitive, in this case, information before returning it to, in this case, a consumer. Useful when the risk is accidental disclosure, at this layer, in model output. Need: apply a GenAI-boundary filter to sensitive output. Post-output PII filter solves another problem.
Question 8
Prompt and response archives in S3 must expire automatically after 30 days under the approved retention policy. Which option most directly addresses the requirement? Choose ONE.
- Short retention plus explicit deletion
- S3 Lifecycle retention policy
- AWS KMS encryption
- CloudWatch access monitoring
Correct Answer: B
Correct Answer
Answer B is correct because S3 Lifecycle retention policy — for this design, transitions or expires S3 objects, for this workload, according to age-based lifecycle rules. Fits when stored prompts, outputs, or evidence, for the stated need, must be deleted or tiered, for the stated need, after a defined retention period. Constraint: automate age-based expiration of stored S3 interaction artifacts. S3 Lifecycle retention policy is the closest fit.
Incorrect Answers
Answer A is incorrect because Short retention plus explicit deletion — in this scenario, keeps interaction artifacts only for, in this case, the approved period and deletes, in this case, them when the retention obligation, in this case, ends. Useful when privacy policy requires reducing the, at this layer, lifetime of stored prompts or, at this layer, outputs. Need: automate age-based expiration of stored S3 interaction artifacts. Short retention plus explicit deletion solves another problem.
Answer C is incorrect because AWS KMS encryption — at this decision point, uses customer-managed or AWS managed, within this design, cryptographic keys to protect supported, within this design, data at rest and control, within this design, key usage. Typical use: the primary control is encryption, in the described flow, and key governance for stored, in the described flow, data. Goal: automate age-based expiration of stored S3 interaction artifacts. AWS KMS encryption does not provide it.
Answer D is incorrect because CloudWatch access monitoring — for this design, collects operational telemetry and alarms, for this workload, so suspicious or unexpected access, for this workload, behavior can be detected. Fits when security teams need monitoring rather, for the stated need, than a preventive data transformation. Constraint: automate age-based expiration of stored S3 interaction artifacts. CloudWatch access monitoring targets another layer.
Question 9
A customer-support prompt needs the issue description but does not need the customer’s full credit-card number. Which design best satisfies this requirement? Choose ONE.
- TLS for data in transit
- Pseudonymization/tokenization
- Anonymization
- Data masking before model invocation
Correct Answer: D
Correct Answer
Answer D is correct because Data masking before model invocation — for this design, replaces sensitive values with masked, for this workload, representations before they are sent, for this workload, to the model. Fits when the model does not need, for the stated need, the original sensitive value to, for the stated need, perform the task. Constraint: remove unnecessary sensitive values before the model sees them. Data masking before model invocation is the closest fit.
Incorrect Answers
Answer A is incorrect because TLS for data in transit — for this design, protects data while it moves, for this workload, between clients and services using, for this workload, encrypted transport. Fits when the threat concerns interception of, for the stated need, traffic in transit. Constraint: remove unnecessary sensitive values before the model sees them. TLS for data in transit targets another layer.
Answer B is incorrect because Pseudonymization/tokenization — in this scenario, replaces direct identifiers with controlled, in this case, tokens so records remain linkable, in this case, without exposing the original identifier. Useful when analytics needs stable linkage while, at this layer, direct identifiers remain protected. Need: remove unnecessary sensitive values before the model sees them. Pseudonymization/tokenization solves another problem.
Answer C is incorrect because Anonymization — in this scenario, removes or transforms identifying attributes, in this case, so the resulting data is, in this case, not intended to be linked, in this case, back to an individual. Useful when the use case does not, at this layer, require re-identification or stable subject, at this layer, linkage. Need: remove unnecessary sensitive values before the model sees them. Anonymization solves another problem.
Question 10
Analytics must correlate repeated interactions from the same user while hiding the user’s direct identifier from analysts. What is the most appropriate implementation choice? Choose ONE.
- Pseudonymization/tokenization
- Data minimization
- Data masking before model invocation
- Anonymization
Correct Answer: A
Correct Answer
Answer A is correct because Pseudonymization/tokenization — for this design, replaces direct identifiers with controlled, for this workload, tokens so records remain linkable, for this workload, without exposing the original identifier. Fits when analytics needs stable linkage while, for the stated need, direct identifiers remain protected. Constraint: preserve stable linkage without exposing the original identifier. Pseudonymization/tokenization is the closest fit.
Incorrect Answers
Answer B is incorrect because Data minimization — for this design, limits collection and prompt context, for this workload, to only the data required, for this workload, for the stated purpose. Fits when privacy risk should be reduced, for the stated need, before applying downstream controls. Constraint: preserve stable linkage without exposing the original identifier. Data minimization targets another layer.
Answer C is incorrect because Data masking before model invocation — at this decision point, replaces sensitive values with masked, within this design, representations before they are sent, within this design, to the model. Typical use: the model does not need, in the described flow, the original sensitive value to, in the described flow, perform the task. Goal: preserve stable linkage without exposing the original identifier. Data masking before model invocation does not provide it.
Answer D is incorrect because Anonymization — for this design, removes or transforms identifying attributes, for this workload, so the resulting data is, for this workload, not intended to be linked, for this workload, back to an individual. Fits when the use case does not, for the stated need, require re-identification or stable subject, for the stated need, linkage. Constraint: preserve stable linkage without exposing the original identifier. Anonymization targets another layer.
Question 11
A public research data set should not permit records to be linked back to individual customers. Which approach should the developer use? Choose ONE.
- Pseudonymization/tokenization
- Short retention plus explicit deletion
- Data masking before model invocation
- Anonymization
Correct Answer: D
Correct Answer
Answer D is correct because Anonymization removes or transforms identifying attributes so the released data is not intended to be linked back to a person. Because the research data set does not require later re-identification, eliminating that linkage is the direct privacy control. Pseudonymization would deliberately retain a reversible or stable linkage mechanism.
Incorrect Answers
Answer A is incorrect because Pseudonymization/tokenization — at this decision point, replaces direct identifiers with controlled, within this design, tokens so records remain linkable, within this design, without exposing the original identifier. Typical use: analytics needs stable linkage while, in the described flow, direct identifiers remain protected. Goal: remove identifying linkage for a data set that does not need re-identification. Pseudonymization/tokenization does not provide it.
Answer B is incorrect because Short retention and explicit deletion reduce how long stored data remains available, which is useful for privacy and compliance. They do not remove identifying linkage from records that are intentionally published. A public research data set must be anonymized before release if individual customers must not be re-identifiable.
Answer C is incorrect because Masking before model invocation hides sensitive values from the model while preserving enough information for an inference task. That protects a processing step, but it does not guarantee that a public research data set cannot be linked back to customers. The release requirement calls for anonymization instead.
Question 12
A vector index stored on an AWS service requires encryption at rest with centrally governed cryptographic keys. What should the team implement? Choose ONE.
- TLS for data in transit
- S3 Block Public Access
- AWS KMS encryption
- Secrets Manager for credentials
Correct Answer: C
Correct Answer
Answer C is correct because AWS KMS encryption — under these constraints, uses customer-managed or AWS managed, at this layer, cryptographic keys to protect supported, at this layer, data at rest and control, at this layer, key usage. Choose it when the primary control is encryption, under these constraints, and key governance for stored, under these constraints, data. Requirement: protect stored data through governed encryption keys. AWS KMS encryption addresses it precisely.
Incorrect Answers
Answer A is incorrect because TLS for data in transit — at this decision point, protects data while it moves, within this design, between clients and services using, within this design, encrypted transport. Typical use: the threat concerns interception of, in the described flow, traffic in transit. Goal: protect stored data through governed encryption keys. TLS for data in transit does not provide it.
Answer B is incorrect because S3 Block Public Access — for this design, prevents public exposure paths for, for this workload, S3 buckets and objects at, for this workload, the account or bucket level. Fits when the key risk is accidental, for the stated need, public access to S3 data. Constraint: protect stored data through governed encryption keys. S3 Block Public Access targets another layer.
Answer D is incorrect because Secrets Manager for credentials — in this scenario, stores and retrieves application secrets, in this case, without hardcoding them into prompts,, in this case, code, or configuration files. Useful when the application must protect rotating, at this layer, credentials used by integrations. Need: protect stored data through governed encryption keys. Secrets Manager for credentials solves another problem.
Question 13
An application uses an external API key for a retrieval source. The key must not appear in code, prompts, or plain environment files. Which option most directly addresses the requirement? Choose ONE.
- Secrets Manager for credentials
- AWS KMS encryption
- Cross-account IAM role
- IAM resource-scoped policy
Correct Answer: A
Correct Answer
Answer A is correct because Secrets Manager for credentials — for this design, stores and retrieves application secrets, for this workload, without hardcoding them into prompts,, for this workload, code, or configuration files. Fits when the application must protect rotating, for the stated need, credentials used by integrations. Constraint: securely store and retrieve a rotating application secret. Secrets Manager for credentials is the closest fit.
Incorrect Answers
Answer B is incorrect because AWS KMS supplies cryptographic key management and can protect data at rest, but it is not the application-facing store for retrieving and rotating an external API credential. AWS Secrets Manager is designed to store, access, and rotate secrets while avoiding credentials in source code, prompts, or plain environment files.
Answer C is incorrect because Cross-account IAM role — for this design, provides temporary, scoped cross-account access, for this workload, without sharing long-lived user credentials. Fits when a workload in one AWS, for the stated need, account must access approved resources, for the stated need, in another. Constraint: securely store and retrieve a rotating application secret. Cross-account IAM role targets another layer.
Answer D is incorrect because IAM resource-scoped policy — under these constraints, grants only the required actions, at this layer, against the specific model, data,, at this layer, or application resources that need, at this layer, access. Choose it when the requirement is authorization minimization, under these constraints, rather than network isolation. Requirement: securely store and retrieve a rotating application secret. IAM resource-scoped policy is not the direct fit.
Question 14
A client sends prompts to an application endpoint across the network; interception of those prompts is the primary threat. Which design best satisfies this requirement? Choose ONE.
- AWS KMS encryption
- VPC endpoint for private service access
- TLS for data in transit
- Data masking before model invocation
Correct Answer: C
Correct Answer
Answer C is correct because TLS for data in transit — under these constraints, protects data while it moves, at this layer, between clients and services using, at this layer, encrypted transport. Choose it when the threat concerns interception of, under these constraints, traffic in transit. Requirement: protect sensitive traffic while it is moving. TLS for data in transit addresses it precisely.
Incorrect Answers
Answer A is incorrect because AWS KMS encryption — in the current architecture, uses customer-managed or AWS managed, in the described flow, cryptographic keys to protect supported, in the described flow, data at rest and control, in the described flow, key usage. Applies where the primary control is encryption, in this case, and key governance for stored, in this case, data. Target outcome: protect sensitive traffic while it is moving. AWS KMS encryption misses that outcome.
Answer B is incorrect because VPC endpoint for private service access — at this decision point, keeps supported AWS service traffic, within this design, on private connectivity without requiring, within this design, public internet traversal. Typical use: a protected GenAI environment must, in the described flow, avoid public network paths. Goal: protect sensitive traffic while it is moving. VPC endpoint for private service access does not provide it.
Answer D is incorrect because Data masking before model invocation — for the described workload, replaces sensitive values with masked, for the stated need, representations before they are sent, for the stated need, to the model. Best suited when the model does not need, for this decision, the original sensitive value to, for this decision, perform the task. Deciding factor: protect sensitive traffic while it is moving. Data masking before model invocation addresses something else.
Question 15
A source bucket containing proprietary documents must never become publicly readable because of an ACL or bucket configuration mistake. What is the most appropriate implementation choice? Choose ONE.
- Amazon Macie sensitive-data discovery
- AWS KMS encryption
- Least-privilege bucket policy
- S3 Block Public Access
Correct Answer: D
Correct Answer
Answer D is correct because S3 Block Public Access — at this decision point, prevents public exposure paths for, within this design, S3 buckets and objects at, within this design, the account or bucket level. Typical use: the key risk is accidental, in the described flow, public access to S3 data. Goal: prevent public S3 exposure paths. S3 Block Public Access provides that behavior.
Incorrect Answers
Answer A is incorrect because Amazon Macie sensitive-data discovery — under these constraints, discovers and classifies sensitive data, at this layer, such as PII in Amazon, at this layer, S3 using managed data-security capabilities. Choose it when the organization must locate sensitive, under these constraints, information stored in S3 at, under these constraints, scale. Requirement: prevent public S3 exposure paths. Amazon Macie sensitive-data discovery is not the direct fit.
Answer B is incorrect because AWS KMS encryption — for this operational need, uses customer-managed or AWS managed, under these constraints, cryptographic keys to protect supported, under these constraints, data at rest and control, under these constraints, key usage. Architects use it when the primary control is encryption, for this workload, and key governance for stored, for this workload, data. Key need: prevent public S3 exposure paths. AWS KMS encryption does not satisfy it.
Answer C is incorrect because Least-privilege bucket policy — in this scenario, restricts S3 object access to, in this case, the exact principals, prefixes, conditions,, in this case, or transport requirements needed. Useful when the risk is overly broad, at this layer, access to source documents. Need: prevent public S3 exposure paths. Least-privilege bucket policy solves another problem.
Question 16
A VPC endpoint exists, but only the application security group should be able to reach the private endpoint ENIs. Which approach should the developer use? Choose ONE.
- TLS for data in transit
- VPC endpoint for private service access
- Security groups for endpoint access
- IAM resource-scoped policy
Correct Answer: C
Correct Answer
Answer C is correct because Security groups for endpoint access — at this decision point, restrict network flows to the, within this design, intended sources and destinations around, within this design, VPC resources or endpoints. Typical use: private connectivity exists but network-level, in the described flow, reachability still needs narrowing. Goal: narrow network reachability around private connectivity. Security groups for endpoint access provides that behavior.
Incorrect Answers
Answer A is incorrect because TLS for data in transit — for the described workload, protects data while it moves, for the stated need, between clients and services using, for the stated need, encrypted transport. Best suited when the threat concerns interception of, for this decision, traffic in transit. Deciding factor: narrow network reachability around private connectivity. TLS for data in transit addresses something else.
Answer B is incorrect because VPC endpoint for private service access — under these constraints, keeps supported AWS service traffic, at this layer, on private connectivity without requiring, at this layer, public internet traversal. Choose it when a protected GenAI environment must, under these constraints, avoid public network paths. Requirement: narrow network reachability around private connectivity. VPC endpoint for private service access is not the direct fit.
Answer D is incorrect because IAM resource-scoped policy — for the described workload, grants only the required actions, for the stated need, against the specific model, data,, for the stated need, or application resources that need, for the stated need, access. Best suited when the requirement is authorization minimization, for this decision, rather than network isolation. Deciding factor: narrow network reachability around private connectivity. IAM resource-scoped policy addresses something else.
Question 17
Sensitive embeddings are encrypted with a customer-managed KMS key, but only the retrieval service role should be able to decrypt them. What should the team implement? Choose ONE.
- Cross-account IAM role
- AWS KMS encryption
- KMS key policy and grants
- IAM resource-scoped policy
Correct Answer: C
Correct Answer
Answer C is correct because KMS key policy and grants — in this scenario, controls which principals can use, in this case, a KMS key for cryptographic, in this case, operations. Useful when encrypted data exists but unauthorized, at this layer, key use must be prevented. Need: control which principal may use the encryption key. KMS key policy and grants matches it directly.
Incorrect Answers
Answer A is incorrect because Cross-account IAM role — at this decision point, provides temporary, scoped cross-account access, within this design, without sharing long-lived user credentials. Typical use: a workload in one AWS, in the described flow, account must access approved resources, in the described flow, in another. Goal: control which principal may use the encryption key. Cross-account IAM role does not provide it.
Answer B is incorrect because AWS KMS encryption — within this implementation, uses customer-managed or AWS managed, for this decision, cryptographic keys to protect supported, for this decision, data at rest and control, for this decision, key usage. Strong fit when the primary control is encryption, within this design, and key governance for stored, within this design, data. Operational need: control which principal may use the encryption key. AWS KMS encryption is secondary here.
Answer D is incorrect because IAM resource-scoped policy — in the current architecture, grants only the required actions, in the described flow, against the specific model, data,, in the described flow, or application resources that need, in the described flow, access. Applies where the requirement is authorization minimization, in this case, rather than network isolation. Target outcome: control which principal may use the encryption key. IAM resource-scoped policy misses that outcome.
Question 18
A GenAI workload in Account A must read an approved S3 prefix in Account B without creating long-lived IAM users. Which option most directly addresses the requirement? Choose ONE.
- Least-privilege bucket policy
- Cross-account IAM role
- Secrets Manager for credentials
- VPC endpoint for private service access
Correct Answer: B
Correct Answer
Answer B is correct because Cross-account IAM role — under these constraints, provides temporary, scoped cross-account access, at this layer, without sharing long-lived user credentials. Choose it when a workload in one AWS, under these constraints, account must access approved resources, under these constraints, in another. Requirement: provide temporary scoped cross-account access. Cross-account IAM role addresses it precisely.
Incorrect Answers
Answer A is incorrect because Least-privilege bucket policy — for this design, restricts S3 object access to, for this workload, the exact principals, prefixes, conditions,, for this workload, or transport requirements needed. Fits when the risk is overly broad, for the stated need, access to source documents. Constraint: provide temporary scoped cross-account access. Least-privilege bucket policy targets another layer.
Answer C is incorrect because Secrets Manager for credentials — at this decision point, stores and retrieves application secrets, within this design, without hardcoding them into prompts,, within this design, code, or configuration files. Typical use: the application must protect rotating, in the described flow, credentials used by integrations. Goal: provide temporary scoped cross-account access. Secrets Manager for credentials does not provide it.
Answer D is incorrect because VPC endpoint for private service access — for the described workload, keeps supported AWS service traffic, for the stated need, on private connectivity without requiring, for the stated need, public internet traversal. Best suited when a protected GenAI environment must, for this decision, avoid public network paths. Deciding factor: provide temporary scoped cross-account access. VPC endpoint for private service access addresses something else.
Question 19
A support workflow currently sends the full customer profile to the model even though only product tier and issue text are needed. Which design best satisfies this requirement? Choose ONE.
- Anonymization
- Data minimization
- Bedrock Guardrails sensitive-information filter
- Data masking before model invocation
Correct Answer: B
Correct Answer
Answer B is correct because Data minimization — at this decision point, limits collection and prompt context, within this design, to only the data required, within this design, for the stated purpose. Typical use: privacy risk should be reduced, in the described flow, before applying downstream controls. Goal: reduce privacy exposure by sending only purpose-required data. Data minimization provides that behavior.
Incorrect Answers
Answer A is incorrect because Anonymization — under these constraints, removes or transforms identifying attributes, at this layer, so the resulting data is, at this layer, not intended to be linked, at this layer, back to an individual. Choose it when the use case does not, under these constraints, require re-identification or stable subject, under these constraints, linkage. Requirement: reduce privacy exposure by sending only purpose-required data. Anonymization is not the direct fit.
Answer C is incorrect because Bedrock Guardrails sensitive-information filter — under these constraints, can detect and block or, at this layer, mask configured sensitive information in, at this layer, model interactions. Choose it when sensitive data must be filtered, under these constraints, at the GenAI interaction boundary. Requirement: reduce privacy exposure by sending only purpose-required data. Bedrock Guardrails sensitive-information filter is not the direct fit.
Answer D is incorrect because Data masking before model invocation — in the current architecture, replaces sensitive values with masked, in the described flow, representations before they are sent, in the described flow, to the model. Applies where the model does not need, in this case, the original sensitive value to, in this case, perform the task. Target outcome: reduce privacy exposure by sending only purpose-required data. Data masking before model invocation misses that outcome.
Question 20
A preprocessing service detects SSNs in incoming text. The model does not need SSNs to answer, so they should be removed before invocation. What is the most appropriate implementation choice? Choose ONE.
- Pre-prompt PII redaction
- Amazon Comprehend PII detection
- Bedrock Guardrails sensitive-information filter
- Post-output PII filter
Correct Answer: A
Correct Answer
Answer A is correct because Pre-prompt PII redaction — in this scenario, removes detected PII from the, in this case, prompt before model invocation so, in this case, unnecessary sensitive values never reach, in this case, the model. Useful when the model can complete the, at this layer, task without raw PII. Need: remove detected PII before the prompt reaches the model. Pre-prompt PII redaction matches it directly.
Incorrect Answers
Answer B is incorrect because Amazon Comprehend PII detection — under these constraints, detects PII entities in text, at this layer, so an application can identify, at this layer, sensitive text before or during, at this layer, processing. Choose it when the application must find PII, under these constraints, in text content programmatically. Requirement: remove detected PII before the prompt reaches the model. Amazon Comprehend PII detection is not the direct fit.
Answer C is incorrect because Bedrock Guardrails sensitive-information filter — for the described workload, can detect and block or, for the stated need, mask configured sensitive information in, for the stated need, model interactions. Best suited when sensitive data must be filtered, for this decision, at the GenAI interaction boundary. Deciding factor: remove detected PII before the prompt reaches the model. Bedrock Guardrails sensitive-information filter addresses something else.
Answer D is incorrect because Post-output PII filter — for this design, checks generated content for sensitive, for this workload, information before returning it to, for this workload, a consumer. Fits when the risk is accidental disclosure, for the stated need, in model output. Constraint: remove detected PII before the prompt reaches the model. Post-output PII filter targets another layer.
