100% Latest & Updated Google Professional Cloud Architect Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 210
A generative-AI assistant accepts customer prompts and third-party retrieval documents that may contain hostile instructions. Which implementation choice most directly delivers defense against instruction injection?
- Put user authorization policy only into retrieved instruction text
- Give agent tools unrestricted production permissions to avoid failed prompts
- Accept all model-produced tool arguments without server-side validation
- Separate untrusted retrieved text from system instructions and enforce tool permissions externally
Explanation
Correct Answer
D:A retrieved document must not gain the privileges of an application control instruction
Incorrect Answers
A:Documents are not trusted access controls. A retrieved document must not gain the privileges of an application control instruction.
B:Expands prompt-injection impact. A retrieved document must not gain the privileges of an application control instruction.
C:Delegates authorization to untrusted output. A retrieved document must not gain the privileges of an application control instruction.
Question 211
A generative-AI assistant accepts customer prompts and third-party retrieval documents that may contain hostile instructions. Which validation would best establish adversarial safety acceptance?
- Test only a polite benign user question
- Measure generated token speed rather than access-control failures
- Check that model endpoint returns 200 responses
- Test injected documents that request forbidden tool actions and verify refusal and audit evidence
Explanation
Correct Answer
D:A clean benchmark prompt does not exercise tool-use escalation through untrusted context
Incorrect Answers
A:Does not exercise malicious retrieval. A clean benchmark prompt does not exercise tool-use escalation through untrusted context.
B:Not adversarial safety evidence. A clean benchmark prompt does not exercise tool-use escalation through untrusted context.
C:Does not prove blocked tool actions. A clean benchmark prompt does not exercise tool-use escalation through untrusted context.
Question 212
A generative-AI assistant accepts customer prompts and third-party retrieval documents that may contain hostile instructions. Which operational action best maintains continuous agent security assurance after launch?
- Approve model releases without rerunning injection tests
- Monitor only inference cost and ignore tool invocation logs
- Monitor prompt injection failures and update evaluations with real attack patterns
- Erase safety violation evidence immediately
Explanation
Correct Answer
C:Model upgrades or retrieval changes can reopen previously mitigated exploit paths
Incorrect Answers
A:Risk can change between model and retriever versions. Model upgrades or retrieval changes can reopen previously mitigated exploit paths.
B:Cannot find privilege misuse. Model upgrades or retrieval changes can reopen previously mitigated exploit paths.
D:Prevents causal incident review. Model upgrades or retrieval changes can reopen previously mitigated exploit paths.
Question 213
A corporate security team must enforce baseline VPC firewall restrictions across hundreds of business-unit projects. Which architecture most directly meets the need for hierarchy-wide network policy?
- Use hierarchical firewall policies for shared security controls with documented exceptions
- Ask each business unit to remember the global firewall baseline manually
- Use public DNS names as the main network access policy
- Grant project owners a full organization network exemption by default
Explanation
Correct Answer
A:Organization and folder-level rules reduce dependence on each project configuring its own baseline
Incorrect Answers
B:Not consistent inheritance. Organization and folder-level rules reduce dependence on each project configuring its own baseline.
C:Names do not enforce traffic restrictions. Organization and folder-level rules reduce dependence on each project configuring its own baseline.
D:Defeats hierarchical governance. Organization and folder-level rules reduce dependence on each project configuring its own baseline.
Question 214
A corporate security team must enforce baseline VPC firewall restrictions across hundreds of business-unit projects. Which implementation choice most directly delivers controlled rule priority?
- Set appropriate firewall priorities and targets with visibility into inherited rule behavior
- Assign broad allow rules higher priority than the intended deny baseline
- Permit every child project to modify inherited firewall policy
- Target every service identity without checking exclusions
Explanation
Correct Answer
A:A higher priority broad allow can defeat lower priority restrictions
Incorrect Answers
B:Overrides restriction. A higher priority broad allow can defeat lower priority restrictions.
C:Breaks central authority. A higher priority broad allow can defeat lower priority restrictions.
D:May interrupt required traffic. A higher priority broad allow can defeat lower priority restrictions.
Question 215
A corporate security team must enforce baseline VPC firewall restrictions across hundreds of business-unit projects. Which validation would best establish proof of inherited network denial?
- Inspect firewall rule count in one existing project
- Test only an allowed flow and skip forbidden traffic
- Test blocked traffic from a newly created child project and inspect effective policy evaluation
- Use a billing report as a proxy for packet enforcement
Explanation
Correct Answer
C:Existing projects being compliant does not demonstrate inheritance for future ones
Incorrect Answers
A:Does not prove child inheritance. Existing projects being compliant does not demonstrate inheritance for future ones.
B:Misses ineffective denies. Existing projects being compliant does not demonstrate inheritance for future ones.
D:Not a firewall test. Existing projects being compliant does not demonstrate inheritance for future ones.
Question 216
A corporate security team must enforce baseline VPC firewall restrictions across hundreds of business-unit projects. Which operational action best maintains review of firewall exceptions after launch?
- Keep old acquired-company exceptions without review
- Audit exception rules and their effective precedence after mergers or subnet changes
- Edit hierarchy priorities ad hoc during incidents without change records
- Ignore deny logs in new projects because source is centrally managed
Explanation
Correct Answer
B:Legacy exceptions and hierarchy changes can silently widen network reachability
Incorrect Answers
A:Legacy rules persist. Legacy exceptions and hierarchy changes can silently widen network reachability.
C:Can introduce new access paths. Legacy exceptions and hierarchy changes can silently widen network reachability.
D:Misses misuse and configuration problems. Legacy exceptions and hierarchy changes can silently widen network reachability.
Question 217
An engineering organization lets one CI identity approve source changes, publish artifacts, and deploy to production. Which architecture most directly meets the need for separation of build and production trust?
- Separate build, attestation, and production-deploy responsibilities through distinct identities
- Allow the builder to approve and deploy its own artifacts without separate checks
- Store approval notes only in the container image label
- Treat a passing vulnerability scan as a substitute for independent approval
Explanation
Correct Answer
A:One compromised identity should not approve its own unreviewed artifact and deploy it
Incorrect Answers
B:Single credential compromise bypasses review. One compromised identity should not approve its own unreviewed artifact and deploy it.
C:Label can be forged with image publishing rights. One compromised identity should not approve its own unreviewed artifact and deploy it.
D:Cannot enforce separation of duties. One compromised identity should not approve its own unreviewed artifact and deploy it.
Question 218
An engineering organization lets one CI identity approve source changes, publish artifacts, and deploy to production. Which implementation choice most directly delivers gated production admission?
- Require independent approval and enforce immutable artifact references in deployment workflow
- Let every developer modify production admission configuration
- Reference container tags that may be moved after review
- Approve deployments by editable comments in the same commit
Explanation
Correct Answer
A:Review gates must be enforced by permissions and policies, not just checkboxes
Incorrect Answers
B:Can bypass gates. Review gates must be enforced by permissions and policies, not just checkboxes.
C:Does not pin approved bytes. Review gates must be enforced by permissions and policies, not just checkboxes.
D:Lacks independent trust boundary. Review gates must be enforced by permissions and policies, not just checkboxes.
Question 219
An engineering organization lets one CI identity approve source changes, publish artifacts, and deploy to production. Which validation would best establish proof of separation of duties?
- Test a reviewed release only without builder-only negative attempts
- Attempt to deploy an unapproved build using only the builder identity
- Count artifact signatures without checking signer authority
- Check deployment latency rather than rejection of unauthorized releases
Explanation
Correct Answer
B:A successful approved release does not prove the builder cannot bypass review
Incorrect Answers
A:Cannot prove a builder is restricted. A successful approved release does not prove the builder cannot bypass review.
C:Signature alone does not prove allowed actor. A successful approved release does not prove the builder cannot bypass review.
D:Misses enforcement. A successful approved release does not prove the builder cannot bypass review.
Question 220
An engineering organization lets one CI identity approve source changes, publish artifacts, and deploy to production. Which operational action best maintains privileged pipeline reviews after launch?
- Leave emergency deployer credentials active indefinitely
- Ignore approval-denial audit logs
- Rotate only application secrets while the build role retains excess permissions
- Periodically inspect deployer roles and approval bypass logs for unauthorized changes
Explanation
Correct Answer
D:Over time emergency access and automation edits can collapse the intended separation
Incorrect Answers
A:Become a permanent bypass. Over time emergency access and automation edits can collapse the intended separation.
B:Hides attempted separation failures. Over time emergency access and automation edits can collapse the intended separation.
C:Does not restore production release trust. Over time emergency access and automation edits can collapse the intended separation.
Google Cloud Certified - Professional Cloud Architect (Professional Cloud Architect) Practice Tests