Claude Certified Associate Governance practice questions
7-day money-back guarantee — full refund within 7 days of purchase if you've completed under 20% of the questions. See pricing →
Certifications Tools Flashcards Career Paths Exam Guides Blog Pricing For Teams About
Try 20 questions free →

Claude Certified Associate Governance, Risk, and Responsible Use: 180 practice questions

Claude Certified Associate 180 questions 12 shown free

12 of the 180 Governance, Risk, and Responsible Use questions in the Certsqill Claude Certified Associate bank, shown in full below. Each one carries an explanation for every option, not just the correct one — the wrong answers are where the marks go.

Preparing for Claude Certified Associate? Take the free 5-min readiness check →

1. Confirm the approved workspace has the required 30-day: Select TWO actions that meet the stated requirement.

Easy
A business analyst must study attendance trends using an approved Claude workspace. Exhibit: the data contains names, learner IDs, campus, week, attendance, and comments; policy permits only aggregate trends, requires a 30-day retention setting, and prohibits unnecessary identifiers. Select TWO actions that meet the stated requirement.

Select two. More than one option is correct — every correct one is ticked below.

  1. Use a personal Claude account, then delete the conversation afterward.
    A personal account does not satisfy the approved-workspace requirement, and deletion does not establish required handling controls.
  2. Confirm the approved workspace has the required 30-day retention setting. ✓
    Verifying the approved workspace and retention requirement aligns processing with organizational handling expectations.
  3. Aggregate attendance to campus-week counts before uploading the data. ✓
    Aggregation supports the stated trend purpose while minimizing individual-level and unnecessary information before processing.
  4. Tell Claude to forget the uploaded records after producing the chart.
    A prompt cannot change contractual retention, training, sharing, or access-control arrangements.
  5. Redact names but upload learner IDs, campus, and comments unchanged.
    Removing names alone leaves direct or indirect identifiers and unnecessary free-text information in the uploaded dataset.
The trap
Treats name removal as complete deidentification and minimization. Assumes natural-language instructions enforce organizational data controls. Confuses later deletion with approved access and retention controls.

Minimize data for the stated purpose and verify approved workspace and retention requirements before processing.

2. Verify figures: Select TWO appropriate approaches.

Medium
A finance operations team uses Claude to prepare a client-facing variance commentary. Company policy requires disclosure when AI materially assists the draft and requires a qualified employee to verify financial claims before release. Select TWO appropriate approaches.

Select two. More than one option is correct — every correct one is ticked below.

  1. Verify figures, assumptions, and interpretations before approving publication. ✓
    Human verification is necessary because generated financial content requires accountable review before consequential external use.
  2. Label the commentary as AI-assisted and identify the accountable reviewer. ✓
    This provides transparency about assistance and identifies human accountability for the released financial communication.
  3. Describe Claude as a certified source of accurate financial conclusions.
    Claude output is not automatically authoritative, accurate, or a substitute for qualified financial review.
  4. Mention AI assistance only if a client directly asks about it.
    The policy requires disclosure proactively when assistance is material, not only after a client inquiry.
  5. Release it without disclosure because an employee edited every paragraph.
    Human editing does not eliminate a policy requirement to disclose material AI assistance.
The trap
Makes required transparency depend on a client’s question. Assumes editing removes the need for transparency. Replaces evidence and accountability with unsupported model authority.

Disclose material assistance and preserve qualified human verification before releasing consequential financial content.

3. Use the approved account: What should the team do?

Hard
A communications team must summarize confidential campaign feedback using only approved Claude accounts and least-privilege access. An approved connector requests more customer access than the task requires. What should the team do?
  1. Use the approved account, minimize the input, and confirm scope. ✓
    This uses approved tooling, limits confidential data, and verifies that access matches the task.
  2. Upload the full export first, then ask the administrator to review connector permissions.
    Permission and minimization checks should occur before unnecessary confidential data is exposed.
  3. Use the connector’s broad access because the final summary will exclude customer names.
    A sanitized output does not justify unnecessary source access or eliminate confidentiality concerns.
  4. Choose an unapproved service because its default permissions appear narrower than those of the approved connector.
    An apparently narrower scope does not override the requirement to use organization-approved tooling.
The trap
Confuses output redaction with least-privilege input access. Delays essential safeguards until after data sharing. Prioritizes apparent access convenience over the stated approval policy.

Use approved tooling, minimize the data, and verify connector scope before proceeding.

4. Authorized copies of the papers or their relevant source: Select TWO inputs that should be obtained first.

Easy
A research coordinator wants Claude to prepare an evidence-grounded grant literature summary. The coordinator provides the topic, audience, five paper titles, and deadline, but no paper contents. Select TWO inputs that should be obtained first.

Select two. More than one option is correct — every correct one is ticked below.

  1. Authorized copies of the papers or their relevant source text. ✓
    Titles identify sources but do not provide the findings needed for a grounded synthesis.
  2. A guarantee that Claude will retain all five papers permanently for future work.
    A retention promise neither supplies source content nor establishes authorized, reliable evidence use.
  3. A commitment to remove citations so the final prose reads more smoothly.
    Removing citations weakens traceability and does not solve the lack of source evidence.
  4. The coordinator’s preferred tone and level of formality.
    Style guidance affects presentation but does not supply evidence for the literature summary.
  5. The specific claims that require support and the expected citation or uncertainty treatment. ✓
    These instructions define what must be supported and how unsupported or uncertain conclusions should be represented.
The trap
Confuses a useful style preference with a required evidence input. Treats less attribution as a solution to missing research material. Treats an unsupported memory expectation as a substitute for source material.

Obtain authorized source material and define how claims, citations, and uncertainty must be handled.

5. Use the minimum aggregate data needed for completion: What should the team do next?

Medium
An education provider has already confirmed its approved Claude workspace, classified learner data, authorization, and retention requirements. The intended task is to identify course-completion trends, and policy does not permit individual learner profiling. What should the team do next?
  1. Keep all records because approved retention permits their use.
    Retention permission does not make every use necessary or appropriate for the stated purpose.
  2. Upload full learner histories and ask Claude to discover useful patterns.
    Broad exploratory access exceeds the stated aggregate purpose and risks creating impermissible individual profiles.
  3. Ask Claude to infer which named learners are likely to fail.
    That request directly conflicts with the stated prohibition on individual learner profiling.
  4. Use the minimum aggregate data needed for completion trends. ✓
    Aggregating to the stated purpose avoids unnecessary individual profiling after the required governance checks are complete.
The trap
Changes an aggregate improvement task into an individual prediction. Lets model curiosity expand the approved purpose and data scope. Confuses permission to retain with permission to use broadly.

Having completed governance checks, minimize the dataset to aggregate information necessary for the stated trend analysis.

6. Compare missed-escalation rates across language groups: Select TWO measures.

Easy
A customer-support team is evaluating a ticket-routing model. Leaders are concerned that customers using different languages may receive less appropriate escalation. They want measures that evaluate the stated outcome across subgroups. Select TWO measures.

Select two. More than one option is correct — every correct one is ticked below.

  1. Compare average response length across language groups.
    Response length does not measure whether tickets received appropriate escalation.
  2. Report only the overall escalation rate for all customers.
    An overall rate can conceal subgroup disparities and cannot answer the stated fairness concern.
  3. Compare missed-escalation rates across language groups. ✓
    Missed escalation is a relevant error outcome for assessing unequal routing performance.
  4. Compare ticket volumes without examining routing outcomes.
    Volume differences do not show whether the model escalated tickets appropriately or equitably.
  5. Compare appropriate-escalation rates across language groups. ✓
    This directly compares the stated routing outcome across the relevant subgroups.
The trap
Aggregates away the subgroup difference being investigated. Measures workload distribution rather than decision quality. Substitutes a writing characteristic for the stated service outcome.

Measure the routing outcome and relevant error rates separately across language subgroups.

7. Limit access to approved aggregate campaign data: What should the team do?

Medium
A marketing team wants Claude to summarize campaign performance. The task needs read-only access to aggregate campaign metrics; it does not need customer-level records or editing rights. A connector requests access to the entire CRM. What should the team do?
  1. Deny every connector, including access to needed campaign aggregates.
    Rejecting all access is unnecessarily restrictive when a narrower approved scope can support the task.
  2. Limit access to approved aggregate campaign data in read-only mode. ✓
    This satisfies the task while applying least privilege to both data scope and permitted actions.
  3. Use customer-level CRM access because campaign metrics may need context.
    Possible future context does not justify access beyond the specifically required aggregate metrics.
  4. Grant full CRM access, then remove it after the summary.
    Temporary broad access still exceeds the task and unnecessarily exposes customer-level records.
The trap
Treats later revocation as justification for excessive initial privilege. Uses speculative convenience to bypass the explicit scope constraint. Confuses least privilege with denying useful, authorized access.

Grant only the approved, read-only aggregate access needed for campaign analysis.

8. Study B further as decision support: Which approach should be pursued?

Medium
A business analyst is assessing two screening approaches for a consequential internal decision. Approach A has 92% overall accuracy and an 8-point subgroup error gap; Approach B has 89% accuracy and a 1-point gap. The organization prioritizes reducing disparity and requires qualified human review. Which approach should be pursued?
  1. Reject both immediately without comparing subgroup results or review controls.
    Immediate rejection prevents informed assessment of the measured tradeoff and available safeguards.
  2. Study B further as decision support, with qualified human review before decisions. ✓
    B better fits the disparity priority, while further validation and human review address its lower overall accuracy and preserve accountability.
  3. Adopt A because it has the higher accuracy.
    This ignores the organization’s stated priority to reduce subgroup disparity.
  4. Select B automatically because its subgroup gap is smaller, without further validation or accountable review.
    The smaller gap is important but does not remove the need for broader validation and qualified human oversight.
The trap
Optimizes aggregate accuracy while overlooking the explicit fairness priority. Avoids evaluating the alternatives instead of managing the stated risk. Treats one favorable metric as sufficient for consequential deployment.

Pursue B as decision support, but validate it further and retain qualified human review.

9. Generalize or suppress indirect identifiers and reassess: What action best meets that requirement?

Medium
An operations team will share support records for an external summary. It removed names and emails but retained exact timestamps, small-area postal codes, rare issue descriptions, and account-tenure details. The team must reduce reidentification risk before sharing. What action best meets that requirement?
  1. Generalize or suppress indirect identifiers and reassess reidentification risk. ✓
    Indirect identifiers can combine to identify people, so they require minimization and a renewed risk assessment.
  2. Replace names with codes and call the dataset anonymous.
    Pseudonymization with codes is not anonymity, especially when quasi-identifiers remain available.
  3. Tell Claude the records are confidential and proceed unchanged.
    A prompt does not remove identifiers, enforce access controls, or alter retention and sharing arrangements.
  4. Share the records because names and emails have already been removed.
    Remaining timestamps, locations, rarity, and tenure may still identify individuals in combination.
The trap
Treats a confidentiality instruction as a technical data safeguard. Confuses pseudonymization with irreversible deidentification. Assumes direct-identifier removal guarantees anonymity.

Address combinations of indirect identifiers, then reassess whether the shared dataset remains reasonably reidentifiable.

10. Keep the remote-worker concern visible: What should the coordinator do?

Medium
An HR coordinator is preparing a rollout recommendation. Exhibit: 92% of managers support rollout, while 8% report errors concentrated among remote workers. The current draft cites only the majority result. What should the coordinator do?
  1. Report only the 92% approval result because it represents the larger group.
    A smaller response group can still reveal a material subgroup problem that decision-makers need to assess.
  2. Keep the remote-worker concern visible. ✓
    Retaining the dissenting evidence prevents a potentially important subgroup issue from being hidden.
  3. Treat the 8% report as conclusive proof that rollout is unsafe without investigating the reported outcomes.
    Preserving dissent does not make it conclusive; the reported remote-worker outcomes still require examination.
  4. Convert both views into one approval percentage so the recommendation has a single result.
    An aggregate percentage can obscure the specific error concern and prevent meaningful subgroup investigation.
The trap
Treats minority evidence as irrelevant because it is less frequent. Collapses relevant dissent into an uninformative average. Overstates preliminary dissent instead of validating its underlying concern.

Preserve the dissent and investigate the reported remote-worker outcomes before recommending rollout.

11. Obtain qualified human review before acting: What should happen next?

Medium
A nonprofit uses Claude to draft recommendations about beneficiary priorities. The draft is coherent, but organizational policy requires qualified human review before any consequential decision. What should happen next?
  1. Obtain qualified human review before acting on the recommendations. ✓
    The required accountable reviewer must assess the recommendations before the nonprofit uses them in a consequential decision.
  2. Show the draft to staff for informal comments.
    General staff feedback does not establish the required qualified review before a consequential decision.
  3. Ask Claude to repeat the analysis.
    A second model response is not independent qualified review and does not satisfy the required decision gate.
  4. Use only recommendations that Claude labels highly confident, then proceed without a formal reviewer.
    Model confidence does not verify evidence, fairness, authorization, or compliance with the required review process.
The trap
More model output does not replace human accountability. Confuses broad feedback with accountable qualified review. Treats confidence language as governance approval.

Obtain qualified human review before using the recommendations.

12. The current approved threshold and effective date: What specific input is missing?

Medium
An education provider asks Claude to apply a deterministic scholarship rule: flag applicants below the approved household-income threshold. The prompt includes applicant incomes but omits the current threshold and its effective date. What specific input is missing?
  1. A second model’s opinion about each applicant
    Another model cannot determine which threshold the provider has officially approved.
  2. A more persuasive explanation of the rule
    Better wording cannot supply the authoritative threshold needed to apply the deterministic rule.
  3. The current approved threshold and effective date ✓
    The rule cannot be applied reliably without the authoritative threshold and the period in which it applies.
  4. A longer description of each applicant
    Additional applicant detail does not resolve the missing policy value and effective period.
The trap
Independent output does not establish the governing rule. Clarity does not substitute for the missing governing value. More case detail cannot replace an absent rule input.

A deterministic rule needs its authoritative threshold and applicable effective date.

168 more Governance, Risk, and Responsible Use questions

The remaining 168 questions in this domain are part of the full Claude Certified Associate bank — 1197 questions, every option explained. Start with the free five-minute check and see your score per domain.

Test your Claude Certified Associate readiness — free

Other Claude Certified Associate domains

Part of the Certsqill Claude Certified Associate question bank · Governance, Risk, and Responsible Use · Every answer, right and wrong, comes with its own explanation.