Question 190
A compliance assurance implementation team has approved Untrusted retrieved-content isolation. Its expected behavior is that the control layer can treat retrieved text as evidence and prohibit it from changing tool policy. What prerequisite belongs on the release-readiness checklist?
- Confirm layered controls and assigned incident owners
- Confirm parsing boundary and policy enforcement outside the model
- Confirm recipient entitlement map and redaction tests
- Confirm central policy service and auditable subject identity
Explanation
Correct Answer
B:Before deployment the team must confirm parsing boundary and policy enforcement outside the model. Without that preparation, the claim that the control layer can treat retrieved text as evidence and prohibit it from changing tool policy may remain unenforceable; the related exposure is that instruction blending permits external authors to steer agent behavior. The acceptance indicator is policy override rate in hostile-document evaluations.
Incorrect Answers
A:This prerequisite supports a method used when failures may bypass one input or model safety layer, because the control layer can combine prevention, detection, containment and audited recovery; however, the named intervention needs parsing boundary and policy enforcement outside the model before its protection can be enforced.
C:Where generated summaries may echo sensitive identifiers unnecessary for recipients, the control layer can enforce recipient-specific redaction and validated output fields; for this release, the team must first verify parsing boundary and policy enforcement outside the model.
D:For the separate challenge where assistants propose privileged write operations in customer systems, the control layer can require server-side authorization and business-policy checks before each side effect; however, the named intervention needs parsing boundary and policy enforcement outside the model before its protection can be enforced.
Question 191
After installing Calibration and abstention quality in an education services workflow, the team must validate whether it has reduced the exposure that accuracy on answered cases alone can reward refusing everything. Which measurement is most relevant?
- Measure total spend divided by validated resolved tickets
- Measure high-percentile latency and timeout frequency
- Measure unauthorized side effects per adversarial test run
- Measure risk-coverage curve and improper certainty rate
Explanation
Correct Answer
D:The correct signal is risk-coverage curve and improper certainty rate; it must be computed for representative workloads and compared with a baseline. It checks whether the evaluation process can evaluate correctness jointly with justified abstentions across uncertainty levels while surfacing the exposure that accuracy on answered cases alone can reward refusing everything.
Incorrect Answers
A:The proposed measure would suit a setting where model usage drops while human rework unexpectedly grows, since the evaluation process can include review, retries and inference in outcome-based unit economics; by contrast, the research administration release needs evidence from risk-coverage curve and improper certainty rate.
B:Where interactive sessions frequently exceed their response deadline, a relevant check would establish whether the evaluation process can track p95 and p99 end-to-end latency by request class; however, that proxy would not demonstrate risk-coverage curve and improper certainty rate for the selected intervention.
C:When an assistant can invoke sensitive operations through tools, the evaluation process can measure disallowed actions that reach a protected side-effecting endpoint; but the acceptance gate must instead inspect risk-coverage curve and improper certainty rate to detect the targeted failure.
Question 192
A public-sector casework service owner is reviewing Value-to-risk prioritization after a canary rollout. The critical regression would occur if prioritizing solely by raw request volume ignores potential harm. What observable acceptance signal belongs in the dashboard?
- Evaluate risk-adjusted impact and implementation effort per use case
- Evaluate change in qualified escalations per thousand completed tasks
- Evaluate end-to-end labor minutes per accepted outcome
- Evaluate reviewer override and acceptance rates by decision category
Explanation
Correct Answer
A:A defensible evaluation looks at risk-adjusted impact and implementation effort per use case. The control's causal claim is that a scored decision framework surfaces high-value lower-risk starting points. Monitoring that outcome, not merely elapsed tokens or success-looking text, helps catch prioritizing solely by raw request volume ignores potential harm.
Incorrect Answers
B:The proposed measure would suit a setting where stakeholders disagree whether faster drafting or fewer escalations is the goal, since a measurable business outcome narrows the solution boundary before model selection; but the acceptance gate must instead inspect risk-adjusted impact and implementation effort per use case to detect the targeted failure.
C:Where executives claim the assistant will save time without measuring current work, a relevant check would establish whether measuring existing effort establishes a credible incremental-value estimate; by contrast, the public-sector casework release needs evidence from risk-adjusted impact and implementation effort per use case.
D:Where caseworkers must retain authority for consequential final decisions, a relevant check would establish whether Claude drafts ranked recommendations while a designated reviewer authorizes each action; however, that proxy would not demonstrate risk-adjusted impact and implementation effort per use case for the selected intervention.
Question 193
The supply-chain planning release team needs a numerical gate for Short-lived scoped service credentials. Rather than equating deployment with success, it wants to catch cases in which long-lived broad secrets amplify compromise impact. Which indicator should it track?
- Evaluate credential scope violations and secret age
- Evaluate forged and replayed webhook rejection rate
- Evaluate cross-tenant unauthorized-read attempts successfully denied
- Evaluate revoked-token access denial and scope mismatch rate
Explanation
Correct Answer
A:The release should evaluate credential scope violations and secret age rather than equate successful calls with real outcomes. The control works on the premise that the integration can issue narrowly scoped renewable credentials for the job identity; the proposed measure can expose the risk that long-lived broad secrets amplify compromise impact.
Incorrect Answers
B:When an upstream application posts actions to a model-driven workflow, the integration can verify message authenticity, timestamp and replay protection before accepting commands; by contrast, the research administration release needs evidence from credential scope violations and secret age.
C:The alternative indicator checks whether the integration can enforce each call using the requesting user's authorized scope when a shared integration must reflect each employee's actual record rights; however, that proxy would not demonstrate credential scope violations and secret age for the selected intervention.
D:The proposed measure would suit a setting where employees connect personal enterprise resources through an integration, since the integration can obtain user consent for minimum scopes and handle revocation; but the acceptance gate must instead inspect credential scope violations and secret age to detect the targeted failure.
Question 194
The post-incident analysis for education services concludes that ambient local settings conceal incompatibilities in shared builds. Which remediation targets the relevant failure boundary?
- Introduce Shared MCP connector governance
- Introduce Reproducible development environment
- Introduce Hook-based policy checks
- Introduce Workspace separation for sensitive tasks
Explanation
Correct Answer
B:This intervention addresses the failure at its source because the engineering workflow can pin dependencies, tooling and environment assumptions across contributors. To keep the issue from recurring, establish versioned setup scripts and build reproducibility checks and verify agent-generated changes passing clean-environment builds; otherwise ambient local settings conceal incompatibilities in shared builds can persist.
Incorrect Answers
A:When ad hoc personal connectors produce inconsistent and insecure access, the engineering workflow can register reviewed MCP servers with clear ownership and allowed scopes; but this remedy operates on another boundary, not the case where ambient local settings conceal incompatibilities in shared builds.
C:Although the alternative reduces the risk that natural-language reminders cannot substitute for enforceable test execution because the engineering workflow can run deterministic checks at controlled workflow boundaries; yet it would not correct the observed weakness: ambient local settings conceal incompatibilities in shared builds.
D:This addresses the failure where a shared default environment increases cross-project exposure, because the engineering workflow can isolate project roots, secrets and tool permissions by approved workspace; but this remedy operates on another boundary, not the case where ambient local settings conceal incompatibilities in shared builds.
Question 195
A design authority for customer success has established this requirement: privacy-sensitive and low-risk workloads require different operating controls. Which intervention should be prioritized?
- Reasoning-capable model for complex synthesis
- Context-capacity fit before model selection
- Per-class model policy
- Empirical model bake-off
Explanation
Correct Answer
C:The architectural distinction is that the model configuration can set explicit allowed model configurations by data class and purpose. The decision should be implemented with approved data classification and deployment policy and evaluated against policy-compliant invocation rate by workload class; otherwise, a single default model policy may violate different governance requirements.
Incorrect Answers
A:For a workload where multi-document arbitration requires several interdependent inferences, a stronger model can justify tradeoffs across heterogeneous evidence; that does not provide the capability actually needed: the model configuration can set explicit allowed model configurations by data class and purpose.
B:When legal analysis requests require unusually long source packets, the model configuration can test effective context utilization and recall before choosing a model configuration; the alternative cannot replace this key requirement: the model configuration can set explicit allowed model configurations by data class and purpose.
D:When stakeholders choose a model based on brand rather than target workload, the model configuration can evaluate candidate models on identical task data and operating constraints; that does not provide the capability actually needed: the model configuration can set explicit allowed model configurations by data class and purpose.
Question 196
A field-service coordination pilot has enabled Data readiness assessment. Because a strong model cannot compensate for absent or unauthorized source data, the release board wants evidence of effectiveness, not just a favorable demonstration. Which metric most directly tests the proposed control?
- Measure outstanding decision-owner ambiguities before design signoff
- Measure agreed priority scores and expected benefits for shortlisted uses
- Measure proportion of nonfunctional requirements with acceptance tests
- Measure discovered exceptions covered by the agreed requirements
- Measure source coverage and freshness pass rate on sampled queries
Explanation
Correct Answer
E:Tracking source coverage and freshness pass rate on sampled queries provides direct evidence that the system delivers its intended effect: the delivery team can validate quality, permissions, update cadence and ownership before promising RAG results. The measure is more informative than prompt counts because the failure being measured is that a strong model cannot compensate for absent or unauthorized source data.
Incorrect Answers
A:The alternative indicator checks whether the delivery team can define accountable decision owners and consultation rights early when product, security and legal teams disagree on who can approve release; but the acceptance gate must instead inspect source coverage and freshness pass rate on sampled queries to detect the targeted failure.
B:When teams propose many attractive pilots without common selection criteria, the delivery team can jointly rank candidates by value, feasibility, harm and data readiness; by contrast, the field-service coordination release needs evidence from source coverage and freshness pass rate on sampled queries.
C:The proposed measure would suit a setting where a proposal lists features but omits response time and recovery requirements, since the delivery team can capture distinct behavior, latency, security, availability and data constraints; however, that proxy would not demonstrate source coverage and freshness pass rate on sampled queries for the selected intervention.
D:The proposed measure would suit a setting where sponsors describe an ideal process that differs from daily operations, since the delivery team can observe actual exception handling and frontline work before mapping automation; but the acceptance gate must instead inspect source coverage and freshness pass rate on sampled queries to detect the targeted failure.
Question 197
The procurement services deployment uses Untrusted retrieved-content isolation, but executive reports currently count only successful model responses. The residual concern is that instruction blending permits external authors to steer agent behavior. Which acceptance measure would reveal real progress?
- Evaluate sensitive data leakage rate in response audits
- Evaluate policy override rate in hostile-document evaluations
- Evaluate prompt-injection success rate against controlled attack fixtures
- Evaluate attack containment time and residual failure severity
- Evaluate cross-tenant fairness and quota bypass incident counts
Explanation
Correct Answer
B:The correct signal is policy override rate in hostile-document evaluations; it must be computed for representative workloads and compared with a baseline. It checks whether the control layer can treat retrieved text as evidence and prohibit it from changing tool policy while surfacing the exposure that instruction blending permits external authors to steer agent behavior.
Incorrect Answers
A:The proposed measure would suit a setting where generated summaries may echo sensitive identifiers unnecessary for recipients, since the control layer can enforce recipient-specific redaction and validated output fields; by contrast, the research administration release needs evidence from policy override rate in hostile-document evaluations.
C:Where user-submitted attachments can contain instructions to steal private data, a relevant check would establish whether the control layer can classify untrusted inputs and prevent them from acquiring instruction authority; but the acceptance gate must instead inspect policy override rate in hostile-document evaluations to detect the targeted failure.
D:When failures may bypass one input or model safety layer, the control layer can combine prevention, detection, containment and audited recovery; by contrast, the research administration release needs evidence from policy override rate in hostile-document evaluations.
E:The alternative indicator checks whether the control layer can enforce bounded usage budgets and anomaly controls per tenant when one customer can monopolize shared model capacity through abusive retries; however, that proxy would not demonstrate policy override rate in hostile-document evaluations for the selected intervention.
Question 198
An independent assessment of a research administration Claude system reports that accuracy on answered cases alone can reward refusing everything. Which change best closes the exposed control gap?
- Deploy Tail-latency measurement
- Deploy Calibration and abstention quality
- Deploy Task-level acceptance rubric
- Deploy Cost per resolved ticket
Explanation
Correct Answer
B:The remediation acts at the relevant design boundary: the evaluation process can evaluate correctness jointly with justified abstentions across uncertainty levels. That counteracts the observed hazard that accuracy on answered cases alone can reward refusing everything. Rollout should include gold labels including unanswerable requests, with regression evidence from risk-coverage curve and improper certainty rate.
Incorrect Answers
A:The alternative reduces the risk that median latency can hide the failures users notice most because the evaluation process can track p95 and p99 end-to-end latency by request class; however, the research administration defect follows the path in which accuracy on answered cases alone can reward refusing everything.
C:This addresses the failure where generic satisfaction scores may reward confident errors, because the evaluation process can define observable criteria for a usable output rather than aesthetic fluency; but this remedy operates on another boundary, not the case where accuracy on answered cases alone can reward refusing everything.
D:Where model usage drops while human rework unexpectedly grows, the evaluation process can include review, retries and inference in outcome-based unit economics; whereas the research administration defect follows the path in which accuracy on answered cases alone can reward refusing everything.
Question 199
Following a regression in field-service coordination, the team establishes that prioritizing solely by raw request volume ignores potential harm. Which proposed redesign most directly prevents recurrence?
- Introduce Outcome-first use-case framing
- Introduce Feasibility pilot with representative cases
- Introduce Value-to-risk prioritization
- Introduce Task-scope exclusion for unsuitable decisions
Explanation
Correct Answer
C:This intervention addresses the failure at its source because a scored decision framework surfaces high-value lower-risk starting points. To keep the issue from recurring, establish a joint rubric agreed with security and operations and verify risk-adjusted impact and implementation effort per use case; otherwise prioritizing solely by raw request volume ignores potential harm can persist.
Incorrect Answers
A:This addresses the failure where feature demonstrations can obscure the absence of an accountable success condition, because a measurable business outcome narrows the solution boundary before model selection; yet it would not correct the observed weakness: prioritizing solely by raw request volume ignores potential harm.
B:When a polished demo can hide unacceptable long-tail failures, small-scope trials expose feasibility and exception handling before a broad commitment; but this remedy operates on another boundary, not the case where prioritizing solely by raw request volume ignores potential harm.
D:Although the alternative reduces the risk that forcing a high-stakes decision into generation creates liability and quality gaps because identifying out-of-scope judgments prevents unsupported automation promises; yet it would not correct the observed weakness: prioritizing solely by raw request volume ignores potential harm.
Question 200
A research administration pilot cannot be accepted because long-lived broad secrets amplify compromise impact. The owner can change one primary architectural control; what should it be?
- Deploy Per-user delegated authorization
- Deploy Short-lived scoped service credentials
- Deploy OAuth consent and delegated scopes
- Deploy Authenticated webhook processing
Explanation
Correct Answer
B:The engineering fix changes the mechanism rather than the presentation: the integration can issue narrowly scoped renewable credentials for the job identity. It requires identity provider integration and secret rotation process; a regression test should examine credential scope violations and secret age in cases where long-lived broad secrets amplify compromise impact.
Incorrect Answers
A:When a shared service token can reveal records across users, the integration can enforce each call using the requesting user's authorized scope; but this remedy operates on another boundary, not the case where long-lived broad secrets amplify compromise impact.
C:When administrator-wide credentials bypass user-controlled consent semantics, the integration can obtain user consent for minimum scopes and handle revocation; yet it would not correct the observed weakness: long-lived broad secrets amplify compromise impact.
D:Although the alternative reduces the risk that trusting an inbound URL lets attackers forge operational events because the integration can verify message authenticity, timestamp and replay protection before accepting commands; but this remedy operates on another boundary, not the case where long-lived broad secrets amplify compromise impact.