A telecommunications operator is choosing between two architectures for a customer-service AI assistant.
Option A retrains a very large foundation model weekly on internal data; Option B fine-tunes a substantially smaller model monthly. Both meet functional accuracy requirements. The AIMS lead is asked how ISO/IEC
42001 treats environmental considerations when an organisation selects an AI design. Which response BEST reflects the standard ' s intent?

• A. Environmental impact should be evaluated by the cloud provider rather than the operator, since the provider operates the underlying infrastructure and is better placed than the operator to compare the relative carbon profiles of competing architectures.
• B. Environmental impact considerations apply solely to AI systems that perform climate-related functions such as emissions forecasting; for a customer service assistant the environmental dimension of architecture choice falls outside the relevant scope.
• C. Environmental considerations sit outside the AIMS and belong with the corporate sustainability team, so the AI architecture decision should be driven purely by functional accuracy and time-to-market without consideration of energy or compute footprint.
• D. Environmental impact is among the considerations the AI System Impact Assessment is expected to address, so the operator should evaluate the energy and compute footprint of each option alongside functional performance before approving an architecture.

A documentation lead is preparing a one-page reference for new joiners that names the four annexes of ISO
/IEC 42001 and their purpose. She wants the description that aligns with GAICC Foundation Module 3.
Which option BEST matches the role of each annex in the standard?

• A. Annex A maps ISO/IEC 42001 to ISO 9001 directly; Annex B is a vocabulary appendix; Annex C is a sample Statement of Applicability; Annex D is a glossary of regional AI regulations.
• B. Annex A is the reference control set; Annex B is implementation guidance; Annex C catalogues AI- related objectives and risk sources; Annex D shows integration with other management system standards.
• C. Annex A is a list of accredited certification bodies; Annex B is a Stage 2 audit checklist; Annex C is a Statement of Applicability template; Annex D is a financial impact assessment model.
• D. Annex A is a published list of certified consultants; Annex B is the official EU AI Act mapping; Annex C is a sample AI policy template; Annex D is the standard ' s revision history log.

An insurer deploys an AI model that supports underwriter decisions on personal-lines policies. Three audiences need explanations: applicants who receive a decline, internal underwriters who supervise the model
' s outputs, and an external regulator conducting a market review. A foundation learner is asked how ISO/IEC
42001 expects explainability to be approached across these audiences. Which response BEST reflects the responsible-AI principle of explainability?

• A. Defer explainability entirely until a complaint or regulatory enquiry is raised, then assemble an explanation reactively for the specific case, since pre-built explanations may not address the particular query the requesting party actually puts to the insurer.
• B. Provide explanations to the regulator and the underwriters but withhold any explanation from declined applicants on commercial-sensitivity grounds, so that proprietary aspects of the underwriting approach remain protected from competitor scrutiny.
• C. Tailor the explanation to each audience: plain-language reasoning for the declined applicant, feature- level rationale for the underwriter, and methodology and validation evidence for the regulator-each calibrated to its purpose.
• D. Provide a single technical explanation document, copied verbatim from the model-development team, to each of the three audiences, on the basis that a uniform explanation supports consistency and auditability across stakeholders interacting with the system.

A small insurance firm is performing its first context analysis under Clause 4.1 of ISO/IEC 42001. The compliance lead lists the factors the team should consider. Which item BEST illustrates an EXTERNAL issue relevant to the AIMS, consistent with GAICC Foundation Module 4?

• A. The firm ' s internal organisational culture and the maturity of its data engineering team, both of which influence how AI projects are scoped and delivered.
• B. The current inventory of AI models deployed in the underwriting platform and the technical debt accumulated within the firm ' s own internal data pipelines over time.
• C. The forthcoming enforcement of regional AI legislation, such as the EU AI Act, and evolving public expectations about transparency in automated insurance decisions.
• D. The composition of the firm ' s senior leadership team and the way responsibilities for AI initiatives are currently allocated across internal business units.

A municipal council trains an AI tool to predict pothole locations using crowdsourced reports submitted through a smartphone app. After deployment, the tool consistently recommends maintenance in wealthier suburbs where smartphone uptake is highest, while under-recommending in lower-income areas where reports are sparse. A foundation learner is asked which type of bias BEST describes what is occurring and which responsible-AI consideration is most directly affected.

• A. Automation bias, because human operators defer to the AI tool ' s recommendations without applying their own knowledge of which neighbourhoods would benefit from inspection regardless of whether residents reported damage through the app.
• B. Confirmation bias, because the council ' s road maintenance team trusts the tool more than it should, and that trust reinforces the existing recommendation pattern through downstream operational scheduling that follows the tool ' s outputs.
• C. Sample selection bias, because the training data systematically over-represents neighbourhoods with higher smartphone uptake and under-represents others, distorting the fairness of the maintenance outcomes the model recommends.
• D. Label bias, because the council misclassified individual reports during the training data labelling stage, and the resulting label errors propagated into the model ' s prediction patterns across the different neighbourhoods served by the council.