Independent subject review and learner calibration are pending. Explanations use an explicit ISO 19011:2018 audit-guidance baseline. ISO 19011:2026 has replaced it; alignment with the current PECB course materials has not yet been verified. The questions are original practice material, and scores are not a validated prediction of your PECB result.
What this exact sample covers
Auditing an AI valuation model
- Stage 1 readiness and sampling
- Nonconformities and certification
- AI-specific controls
- Question 01Domain 4 · evaluation
At an AIMS stage 1 audit, the team finds a documented AI policy, AI objectives, an AI system impact assessment and a Statement of Applicability. Operational controls went live three weeks ago, and no internal audit or management review has yet been planned. The client wants stage 2 to go ahead next week as booked. What should the audit team leader conclude?
- AStage 2 can go ahead as booked, because stage 1 reviews the documented information and the implementation is then verified at stage 2.
- BStage 2 should wait, because stage 1 judges readiness, including whether internal audits and management reviews are planned and done.
- CStage 2 can go ahead once the client submits dates for internal audit and management review, to be verified at first surveillance.
the answer and reasoning
Correct answerB. Stage 2 should wait, because stage 1 judges readiness, including whether internal audits and management reviews are planned and done.
Stage 1 is not only a document review. Under ISO/IEC 17021-1 it evaluates whether the client is ready for stage 2, including whether internal audits and management reviews are being planned and performed and whether the level of implementation supports moving to stage 2. With controls live for three weeks and neither activity even planned, the team leader should report that the client is not ready and that stage 2 should be rescheduled.
Why the other options fail
- A. Stage 1 also judges readiness for stage 2, so verifying implementation later does not remove the need to conclude whether the client is ready now.
- C. Internal audit and management review are evidence of a functioning AIMS before certification; dates alone do not show they have been performed, and they cannot be deferred to surveillance.
Published exam-scope sources: ISO/IEC 17021-1:2015 — Requirements for bodies providing audit and certification of management systems · PECB ISO/IEC 42001 Lead Auditor course · ISO/IEC 42001:2023 — Artificial intelligence management system
Scenario for questions 2–5
Pentreath Valuation: stage 2 audit of a valuation model
Pentreath Valuation Services provides residential property valuations to eleven mortgage lenders. Its automated valuation model (AVM) estimates property values from sales records, property attributes and a flood-risk score bought from a data supplier. Pentreath is undergoing the stage 2 certification audit of its artificial intelligence management system (AIMS) against ISO/IEC 42001:2023; the AIMS covers the development and operation of the AVM.
Pentreath's operating procedure requires every valuation with a model confidence rating of ‘low’ to be held and reviewed by a chartered surveyor before it is released to the lender. In the last quarter the AVM produced 38,400 valuations. The system log records the confidence rating and release time of every valuation, and the surveyor review queue records each completed review.
In July the AVM was extended from urban to rural properties, which now make up 14% of requests. The AI risk assessment was updated in June to cover the extension. The AI system impact assessment for the AVM, which considers the homeowners and borrowers affected by valuations, is dated the previous February and has not been revised. In May the data supplier changed the method behind its flood-risk score; Pentreath's provenance records for the AVM name the supplier and the dataset but not the version used or the date each version was received.
On the second day, the audit team graded one finding as a major nonconformity: no internal audit has yet covered the AVM's operational processes. At the closing meeting, Pentreath's managing director asks the audit team leader to recommend certification now and to have the corrective action verified at the first surveillance audit.
Question 02Domain 4 · analysisThe audit team plans to test whether every low-confidence valuation is reviewed by a surveyor before release. Which sample population would best test this control?
- ACompleted reviews drawn from the surveyor review queue, checking that each records the surveyor's name, the review date and the value released.
- BValuations chosen by Pentreath's head of valuation as typical of the quarter, checking each against a surveyor review and its release time.
- CLow-confidence valuations drawn from the system log, tracing each one to a completed surveyor review dated before the valuation's release.
the answer and reasoning
Correct answerC. Low-confidence valuations drawn from the system log, tracing each one to a completed surveyor review dated before the valuation's release.
To test whether a control was applied to every case it covers, the auditor samples from the full population that should have triggered it and traces each item forward to evidence of the control. Only the system log contains all low-confidence valuations, including any released without review. Sampling the review queue cannot reveal missing reviews, and an auditee-chosen sample is not objective evidence.
Why the other options fail
- A. The review queue contains only valuations that reached a surveyor, so a sample from it cannot detect low-confidence valuations released without any review.
- B. A sample chosen by the auditee can be selected favourably, and typical valuations need not be low-confidence cases at all, so it does not test this control.
Published exam-scope sources: ISO 19011:2018 — Guidelines for auditing management systems (explicit reference baseline) · PECB ISO/IEC 42001 Lead Auditor course
- Question 03Domain 2 · analysis
Based on the evidence about the rural extension, which requirement of ISO/IEC 42001:2023 should the audit team cite as not fulfilled?
- AClause 8.4, because AI system impact assessments are performed when significant changes are proposed, and the assessment predates the extension.
- BClause 6.3, because changes to the AIMS are to be carried out in a planned manner, and the extension lacked a documented AIMS change plan.
- CClause 8.2, because AI risk assessments are performed after each change occurs, and the June update was made before the July extension.
the answer and reasoning
Correct answerA. Clause 8.4, because AI system impact assessments are performed when significant changes are proposed, and the assessment predates the extension.
Clause 8.4 requires AI system impact assessments according to clause 6.1.4 at planned intervals or when significant changes are proposed to occur. Extending the AVM to rural properties changes who is affected and how, yet the impact assessment is still dated February. The risk assessment was updated when the change was proposed, and the change concerned an AI system rather than the AIMS itself.
Why the other options fail
- B. Clause 6.3 addresses changes to the AIMS, whereas the extension changed an AI system; the scenario also gives no evidence that a change plan was missing.
- C. Clause 8.2 is triggered when significant changes are proposed or occur, so updating the risk assessment in June, before the July extension, conforms.
Published exam-scope sources: ISO/IEC 42001:2023 — Artificial intelligence management system · PECB ISO/IEC 42001 Lead Auditor course
- Question 04Domain 6 · evaluation
The managing director asks for a certification recommendation now, with corrective action for the major nonconformity verified at the first surveillance audit. Would agreeing be in accordance with ISO/IEC 17021-1?
- ANo, because for a major nonconformity the correction and corrective action must be reviewed, accepted and verified as effective before certification.
- BYes, because an accepted corrective action plan for a major nonconformity can be verified at the first surveillance audit after certification.
- CYes, because a finding on internal audit coverage does not affect the AVM's outputs, so it can be regraded as minor once a plan is submitted.
the answer and reasoning
Correct answerA. No, because for a major nonconformity the correction and corrective action must be reviewed, accepted and verified as effective before certification.
Under ISO/IEC 17021-1, a certification decision requires the certification body to have reviewed, accepted and verified the effectiveness of correction and corrective action for every major nonconformity; deferral to surveillance applies only to accepted plans for minor nonconformities. If verification is not possible within six months of the last day of stage 2, another stage 2 audit is needed first.
Why the other options fail
- B. Accepting a plan and verifying it at surveillance is the treatment for minor nonconformities; major nonconformities must be verified before certification can be granted.
- C. Grading follows the evidence and its effect on the management system, not the client's response; the absence of internal audit of core processes is not downgraded by submitting a plan.
Published exam-scope sources: ISO/IEC 17021-1:2015 — Requirements for bodies providing audit and certification of management systems · PECB ISO/IEC 42001 Lead Auditor course
- Question 05Domain 5 · evaluation
What should the audit team conclude about Pentreath's implementation of control A.7.5 (data provenance) for the flood-risk score?
- AIt is effectively implemented, because the records identify the supplier and the dataset, and tracking versions is the supplier's responsibility.
- BIt is not effectively implemented, because the records cannot show which version of the score informed valuations before and after May.
- CIt cannot be assessed at this audit, because the supplier's change of method must first be evaluated by the supplier's own certification body.
the answer and reasoning
Correct answerB. It is not effectively implemented, because the records cannot show which version of the score informed valuations before and after May.
Control A.7.5 requires a process for recording the provenance of data used in AI systems over the life cycles of the data and the AI system. Pentreath's records name the source but not the version or receipt date, so it cannot tell which flood-risk method fed any given valuation after the May change. That gap is Pentreath's to close and is auditable now.
Why the other options fail
- A. Naming a supplier and dataset is not provenance over the data life cycle; Pentreath remains responsible for recording which version its own AI system used.
- C. The audit concerns Pentreath's AIMS and its own provenance records, which the team can examine now; a supplier's certification body has no role in that conclusion.
Published exam-scope sources: ISO/IEC 42001:2023 — Artificial intelligence management system · PECB ISO/IEC 42001 Lead Auditor course
Continue preparing
Ready for the full Lead Auditor bank?
Casual mode shows feedback as you answer. Full access adds the complete weighted bank, exam mode, saved results and domain-level analysis for this exact qualification. Start with a 3-day free trial; practice sessions are saved to your account.
Written by Harry Sidhu. The author's credential scope and the source boundary for each standard are published openly.
Compare every ISO/IEC 42001 preparation pathway.