100% Latest & Updated Google Generative AI Leader Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 760
A risky cloud credential is exposed and must be revoked while affected systems are investigated. What practice coordinates these steps?
- Vulnerability assessment
- Incident response process
- Configuration hardening
- Security Command Center
Explanation
Correct Answer
B:Incident response manages containment and recovery. Incident response coordinates containment, investigation and recovery after security events.
Incorrect Answers
A:Vulnerability assessment identifies known weaknesses in components and configurations. This does not directly meet the case requirement to respond methodically to credential exposure.
C:Hardening reduces attack surface through safer settings and permitted services. This does not directly meet the case requirement to respond methodically to credential exposure.
D:Security Command Center centralizes security posture and findings for supported Google Cloud resources. This does not directly meet the case requirement to respond methodically to credential exposure.
Question 761
A supplier uploads an unverified training dataset that may contain malicious examples. Which control should come first?
- Training pipeline protection
- Prompt injection mitigation
- Runtime tool authorization
- Secure data ingestion
Explanation
Correct Answer
D:Source validation reduces poisoning and misuse risks. Secure ingestion checks source provenance, permissions and poisoning risks before using data.
Incorrect Answers
A:Training protection secures datasets, jobs, artifacts and credentials during model development. This does not directly meet the case requirement to screen and authorize data entering the ML pipeline.
B:Injection mitigation blocks attempts to promote external content into privileged instructions. This does not directly meet the case requirement to screen and authorize data entering the ML pipeline.
C:Runtime authorization limits agent operations and third-party tool calls. This does not directly meet the case requirement to screen and authorize data entering the ML pipeline.
Question 762
A training job accesses confidential inputs using shared personal credentials. Which lifecycle control needs improvement?
- Secure data ingestion
- Prompt injection mitigation
- Runtime tool authorization
- Training pipeline protection
Explanation
Correct Answer
D:Training environments and credentials require isolation and secure identities. Training protection secures datasets, jobs, artifacts and credentials during model development.
Incorrect Answers
A:Secure ingestion checks source provenance, permissions and poisoning risks before using data. This does not directly meet the case requirement to protect the training pipeline's access rights.
B:Injection mitigation blocks attempts to promote external content into privileged instructions. This does not directly meet the case requirement to protect the training pipeline's access rights.
C:Runtime authorization limits agent operations and third-party tool calls. This does not directly meet the case requirement to protect the training pipeline's access rights.
Question 763
An organization downloads a model package and needs to ensure it matches the approved release. Which control applies?
- Model artifact integrity
- Continuous patch and monitoring
- Secure data ingestion
- Training pipeline protection
Explanation
Correct Answer
A:Artifact integrity checks hashes, signatures or provenance. Artifact integrity verifies the intended model or package has not been modified unexpectedly.
Incorrect Answers
B:Operations apply patches and watch model-serving systems for new threats. This does not directly meet the case requirement to confirm the deployed model was not tampered with.
C:Secure ingestion checks source provenance, permissions and poisoning risks before using data. This does not directly meet the case requirement to confirm the deployed model was not tampered with.
D:Training protection secures datasets, jobs, artifacts and credentials during model development. This does not directly meet the case requirement to confirm the deployed model was not tampered with.
Question 764
An AI agent uses a tool that can issue payments; finance policy permits only certain amounts. Which technical boundary is vital?
- Training pipeline protection
- Secure data ingestion
- Model artifact integrity
- Runtime tool authorization
Explanation
Correct Answer
D:Tool authorization should enforce allowed operations and limits. Runtime authorization limits agent operations and third-party tool calls.
Incorrect Answers
A:Training protection secures datasets, jobs, artifacts and credentials during model development. This does not directly meet the case requirement to restrict payment actions at runtime.
B:Secure ingestion checks source provenance, permissions and poisoning risks before using data. This does not directly meet the case requirement to restrict payment actions at runtime.
C:Artifact integrity verifies the intended model or package has not been modified unexpectedly. This does not directly meet the case requirement to restrict payment actions at runtime.
Question 765
A search result contains malicious text asking the assistant to ignore all policies. What control is needed?
- Training pipeline protection
- Model artifact integrity
- Prompt injection mitigation
- Continuous patch and monitoring
Explanation
Correct Answer
C:Instruction isolation mitigates prompt injection. Injection mitigation blocks attempts to promote external content into privileged instructions.
Incorrect Answers
A:Training protection secures datasets, jobs, artifacts and credentials during model development. This does not directly meet the case requirement to treat retrieved material as untrusted content.
B:Artifact integrity verifies the intended model or package has not been modified unexpectedly. This does not directly meet the case requirement to treat retrieved material as untrusted content.
D:Operations apply patches and watch model-serving systems for new threats. This does not directly meet the case requirement to treat retrieved material as untrusted content.
Question 766
A production serving dependency receives an urgent security update. Which activity handles this?
- Continuous patch and monitoring
- Prompt injection mitigation
- Model artifact integrity
- Training pipeline protection
Explanation
Correct Answer
A:Patching and monitoring maintain protection after launch. Operations apply patches and watch model-serving systems for new threats.
Incorrect Answers
B:Injection mitigation blocks attempts to promote external content into privileged instructions. This does not directly meet the case requirement to keep runtime components secure over time.
C:Artifact integrity verifies the intended model or package has not been modified unexpectedly. This does not directly meet the case requirement to keep runtime components secure over time.
D:Training protection secures datasets, jobs, artifacts and credentials during model development. This does not directly meet the case requirement to keep runtime components secure over time.
Question 767
A web crawler imports unlicensed and potentially adversarial content into a RAG index. Which stage's security check should be enhanced?
- Model artifact integrity
- Secure data ingestion
- Continuous patch and monitoring
- Training pipeline protection
Explanation
Correct Answer
B:Ingress validation governs data source quality and rights. Secure ingestion checks source provenance, permissions and poisoning risks before using data.
Incorrect Answers
A:Artifact integrity verifies the intended model or package has not been modified unexpectedly. This does not directly meet the case requirement to verify and sanitize incoming source data.
C:Operations apply patches and watch model-serving systems for new threats. This does not directly meet the case requirement to verify and sanitize incoming source data.
D:Training protection secures datasets, jobs, artifacts and credentials during model development. This does not directly meet the case requirement to verify and sanitize incoming source data.
Question 768
A model training pipeline leaves its temporary credentials in a public build log. What lifecycle safeguard failed?
- Model artifact integrity
- Training pipeline protection
- Secure data ingestion
- Runtime tool authorization
Explanation
Correct Answer
B:Secure training operations must protect secrets. Training protection secures datasets, jobs, artifacts and credentials during model development.
Incorrect Answers
A:Artifact integrity verifies the intended model or package has not been modified unexpectedly. This does not directly meet the case requirement to keep build credentials confidential.
C:Secure ingestion checks source provenance, permissions and poisoning risks before using data. This does not directly meet the case requirement to keep build credentials confidential.
D:Runtime authorization limits agent operations and third-party tool calls. This does not directly meet the case requirement to keep build credentials confidential.
Question 769
Security compares deployed model hashes to an approved artifact manifest after an incident. Which control is used?
- Runtime tool authorization
- Prompt injection mitigation
- Model artifact integrity
- Continuous patch and monitoring
Explanation
Correct Answer
C:Integrity attestation detects unauthorized changes. Artifact integrity verifies the intended model or package has not been modified unexpectedly.
Incorrect Answers
A:Runtime authorization limits agent operations and third-party tool calls. This does not directly meet the case requirement to verify trusted model artifacts.
B:Injection mitigation blocks attempts to promote external content into privileged instructions. This does not directly meet the case requirement to verify trusted model artifacts.
D:Operations apply patches and watch model-serving systems for new threats. This does not directly meet the case requirement to verify trusted model artifacts.
Question 770
An internal agent can invoke a data-delete function without approval. What runtime safeguard should apply?
- Prompt injection mitigation
- Training pipeline protection
- Runtime tool authorization
- Secure data ingestion
Explanation
Correct Answer
C:Operational permissions and confirmation gates restrict damaging actions. Runtime authorization limits agent operations and third-party tool calls.
Incorrect Answers
A:Injection mitigation blocks attempts to promote external content into privileged instructions. This does not directly meet the case requirement to enforce allowed destructive tool actions.
B:Training protection secures datasets, jobs, artifacts and credentials during model development. This does not directly meet the case requirement to enforce allowed destructive tool actions.
D:Secure ingestion checks source provenance, permissions and poisoning risks before using data. This does not directly meet the case requirement to enforce allowed destructive tool actions.
Generative AI Leader Practice Tests
