ISO 42001 Lead Auditor cheat sheet
ISO 42001 Lead Auditor · Keep the key ideas within reach.
Fundamental principles and concepts of an AI management system
ISO/IEC 42001:2023 is the first international management system standard for AI and follows the harmonized structure with clauses 4 to 10.
An AI management system (AIMS) is the part of an organization's management system used to plan, implement, maintain, review and improve the management of AI.
An AI system is an engineered system that generates outputs such as predictions, content, recommendations or decisions for a given set of objectives, with varying levels of autonomy.
AI system impact is the effect of an AI system on individuals, groups of people, organizations, society or the environment, and may be positive or negative.
Annex A is normative (reference control objectives and controls); Annexes B, C and D are informative guidance, risk-source examples and sector guidance.
Annex A of ISO/IEC 42001:2023 contains 9 control objectives (A.2 to A.10) and 38 controls covering policy, organization, resources, impacts, life cycle, data, information, use and third parties.
The standard is applicable to any organization, regardless of size, type or sector, that develops, provides or uses AI-based products or services.
The AIMS follows a continual improvement cycle across clauses 4 to 10: context, leadership, planning, support, operation, performance evaluation and improvement.
ISO/IEC 42001 complements rather than replaces ISO/IEC 27001; AI-specific risks such as impact on individuals and unintended behaviour are addressed in addition to information security.
AI management system requirements
Clause 4 requires determining external and internal issues, the needs of interested parties, the AIMS scope, and establishing the AIMS and its processes.
Clause 5 requires top management to establish an AI policy (5.2) and to assign roles, responsibilities and authorities (5.3).
Clause 6.1.2 requires AI risk assessment; 6.1.3 requires AI risk treatment including a Statement of Applicability; 6.1.4 requires an AI system impact assessment.
Clause 6.2 requires measurable AI objectives at relevant functions and levels, together with plans for achieving them.
Clause 7 covers resources, competence, awareness, communication and documented information.
Clause 8 covers operational planning and control and requires the AI risk assessment, risk treatment and impact assessment to be performed operationally (8.2 to 8.4).
Clause 9 requires monitoring, measurement, analysis and evaluation (9.1), internal audit (9.2) and management review (9.3).
Clause 10 requires continual improvement (10.1) and control of nonconformities with corrective action (10.2).
Fundamental audit concepts and principles
An audit is a systematic, independent and documented process for obtaining objective evidence and evaluating it objectively against audit criteria.
ISO 19011 defines seven audit principles: integrity, fair presentation, due professional care, confidentiality, independence, evidence-based approach and risk-based approach.
Audit criteria are the requirements used as a reference (ISO/IEC 42001, legal and regulatory requirements, the organization's own AIMS documents); the scope defines the boundaries of the audit.
First-party audits are internal audits; second-party audits are performed on suppliers or other external providers; third-party audits are independent, such as certification audits.
Audit evidence must be verifiable, recorded and traceable; sampling is used because examining the full population within the audit time is not feasible.
Audit findings are classified as conformity or nonconformity; nonconformities may be graded major or minor, and observations or opportunities for improvement may also be reported.
Auditors must be independent of the activity being audited and must not audit their own work.
Audit conclusions state the extent of conformity with the audit criteria and, where in scope, the effectiveness of the management system.
Preparing an ISO/IEC 42001 audit
An audit programme is the set of audits planned over a time frame; an audit plan covers a single audit and its schedule and arrangements.
Stage 1 of an initial certification audit reviews documented information and readiness; stage 2 evaluates implementation and effectiveness of the AIMS.
Key AIMS documents to review: AIMS scope, AI policy, risk assessment and treatment records, Statement of Applicability, impact assessments, internal audit results and management review outputs.
The audit plan states objectives, criteria, scope, dates, sites, audit team, schedule and the allocation of time per area.
Audit time depends on the AIMS scope and complexity, number and type of AI systems, level of AI risk, number of sites and use of remote auditing techniques.
Sampling should follow risk and the audit trail from the AI risk assessment to implemented Annex A controls and to evidence that the controls operate.
The opening meeting agenda covers introductions, confirmation of plan, scope and criteria, communication routes, working arrangements and classification of findings.
Prepared checklists support but do not replace auditor judgement and the follow-up of leads found during the audit.
Conducting an ISO/IEC 42001 audit
Evidence is collected through interviews, observation of activities and review of documents and records; findings should be corroborated by more than one source.
A written nonconformity states the requirement (audit criteria), the objective evidence, and the gap between the evidence and the requirement.
A major nonconformity means the AIMS cannot achieve its intended results (e.g. missing required process, systemic failure, unaddressed legal requirement); a minor nonconformity is an isolated lapse.
The audit trail links the AI risk assessment to controls listed in the Statement of Applicability and to records demonstrating those controls operate.
For Annex A, auditors verify impact assessment (A.5), AI life cycle controls (A.6), data quality and provenance (A.7), information for interested parties (A.8) and third-party relationships (A.10).
Audit progress is controlled through audit team meetings and communication with the auditee; any change to scope or plan is agreed with the client and recorded.
Auditors use open questions and follow leads; undocumented verbal claims are not accepted as the sole basis for a finding.
Findings are reviewed with the auditee's management before the closing meeting to confirm factual accuracy.
Closing an ISO/IEC 42001 audit
The closing meeting presents findings, nonconformities and conclusions and states the report distribution and follow-up process.
The audit report includes objectives, scope, criteria, team, dates, findings, conclusions and, where applicable, a statement on conformity and effectiveness.
The audit ends when the audit report is issued; defined follow-up activities, where required, are planned and may be conducted as a separate audit.
The auditee responds to nonconformities with correction, root-cause analysis, corrective action and a completion date; the auditor verifies the effectiveness of the action taken.
Certification decisions are made by the certification body independently of the audit team, based on the audit report and related information.
Records of the audit, including findings and reports, are retained as documented information for traceability and confidentiality.
Audit results feed management review and continual improvement under clauses 9.3 and 10.1.
Managing an ISO/IEC 42001 audit programme
An audit programme is a set of one or more audits planned for a specific time frame, with defined objectives, extent, resources and procedures.
Programme planning is risk-based: frequency, duration and depth are driven by AI risk, previous audit results, changes to the organization and complaints.
A certification cycle consists of an initial certification audit (stage 1 and stage 2), surveillance audits at least once a year, and recertification before the three-year cycle expires.
The programme manager identifies risks and opportunities affecting the programme (e.g. auditor competence gaps, scheduling conflicts, auditee availability) and plans mitigation.
Programme resources include auditors competent in AI technology, AI risk and the applicable legal and regulatory context, plus sufficient time for each audit.
Programme performance is monitored and reviewed (e.g. audit completion, trends in findings, timeliness) and the programme is adjusted accordingly.
Records of the programme and of each audit are retained to demonstrate that the programme was implemented as planned.