Zum Hauptinhalt springen
AI Compliance Starter Kit · Register

AI Statement of Applicability (ISO/IEC 42001)

Updated on 1 min Reviewed by: Cenedril Editorial
ISO/IEC 42001 6.1.3 d, Annex A

In an ISO/IEC 42001 certification audit, the Statement of Applicability is the table of contents of the audit: for each control in Annex A it shows whether it applies, why, how far it is implemented and where the evidence is. The template lists all 38 controls with justification, implementation status and reference to the policies of the AI Compliance Starter Kit.

What does the template contain?

The CSV file has nine columns and 38 rows. In the example organisation all controls apply because it buys in systems and develops its own; three controls are partially implemented or planned. The references point to the section numbers of the free policies.

Template: AI Statement of Applicability

Register Template

AI Statement of Applicability

IDControlAnnex AApplicableJustification for inclusionReason for exclusionJustification for exclusionImplementation statusReference
A.2.2AI policyA.2YesGoverning document of the AI management system (clause 5.2)––ImplementedAI Policy 3
A.2.3Alignment with other organizational policiesA.2YesAI use touches information security and data protection––ImplementedAI Policy 3
A.2.4Review of the AI policyA.2YesPolicy must stay current with new systems and legal changes––ImplementedAI Policy 24
A.3.2AI roles and responsibilitiesA.3YesResponsibility for operation, oversight and in-house development must be assigned––ImplementedAI Policy 16
A.3.3Reporting of concernsA.3YesEmployees and third parties need a channel for concerns––ImplementedAI Policy 17
A.4.2Resource documentationA.4YesResources per system must be known for operation and dependencies––ImplementedAI Policy 9; AI register
A.4.3Data resourcesA.4YesOwn systems are trained on own data––ImplementedAI Policy 9; AI Provider Policy 15
A.4.4Tooling resourcesA.4YesThird-party tools and models are part of the systems––ImplementedAI Policy 9
A.4.5System and computing resourcesA.4YesOperation in the cloud and in the own data centre––ImplementedAI Policy 9; AI register
A.4.6Human resourcesA.4YesCompetence for use, oversight and development––ImplementedAI Policy 10
A.5.2AI system impact assessment processA.5YesSystems affect applicants, customers and insured persons––ImplementedAI Policy 8
A.5.3Documentation of AI system impact assessmentsA.5YesResults must be evidenced and retained––ImplementedAI Policy 8; Fundamental rights impact assessment
A.5.4Assessing AI system impact on individuals or groups of individualsA.5YesHigh-risk systems affect vulnerable groups––ImplementedAI Policy 8; Fundamental rights impact assessment
A.5.5Assessing societal impacts of AI systemsA.5YesSocietal impact of the insurance system––Partially implementedAI Policy 8; Fundamental rights impact assessment
A.6.1.2Objectives for responsible development of AI systemsA.6YesIn-house development of two systems––ImplementedAI Policy 5; AI Provider Policy 12
A.6.1.3Processes for responsible AI system design and developmentA.6YesIn-house development with a life cycle––ImplementedAI Policy 15; AI Provider Policy 8
A.6.2.2AI system requirements and specificationA.6YesRequirements of own systems––ImplementedAI Provider Policy 9
A.6.2.3Documentation of AI system design and developmentA.6YesEvidence of development for the conformity assessment––ImplementedAI Provider Policy 11
A.6.2.4AI system verification and validationA.6YesRelease only after testing against acceptance thresholds––ImplementedAI Provider Policy 13
A.6.2.5AI system deploymentA.6YesPutting own and bought-in systems into service––ImplementedAI Provider Policy 8
A.6.2.6AI system operation and monitoringA.6YesMonitoring of all systems in operation––ImplementedAI Usage Policy 9; AI Provider Policy 17
A.6.2.7AI system technical documentationA.6YesInstructions for use and technical documentation of own systems––ImplementedAI Provider Policy 6
A.6.2.8AI system recording of event logsA.6YesLogs required for high-risk systems––ImplementedAI Usage Policy 10; AI Provider Policy 21
A.7.2Data for development and enhancement of AI systemA.7YesOwn systems learn from own data––ImplementedAI Provider Policy 15
A.7.3Acquisition of dataA.7YesData from own holdings and from third parties––ImplementedAI Provider Policy 15
A.7.4Quality of data for AI systemsA.7YesData quality determines fairness of the assessment––ImplementedAI Provider Policy 15
A.7.5Data provenanceA.7YesProvenance per dataset for traceability––Partially implementedAI Provider Policy 15
A.7.6Data preparationA.7YesPreparation influences bias––ImplementedAI Provider Policy 15
A.8.2System documentation and information for usersA.8YesUsers and deployers need information about the system––ImplementedAI Provider Policy 6; AI communication plan
A.8.3External reportingA.8YesReporting channels for third parties on impacts––PlannedAI Policy 17; AI communication plan
A.8.4Communication of incidentsA.8YesIncidents must be communicated internally and to authorities––ImplementedAI Usage Policy 7; AI Provider Policy 18; AI communication plan
A.8.5Information for interested partiesA.8YesAuthorities, customers and affected persons receive information––ImplementedAI Provider Policy 19; AI communication plan
A.9.2Processes for responsible use of AI systemsA.9YesAll employees use AI services––ImplementedAI Usage Policy 4
A.9.3Objectives for responsible use of AI systemsA.9YesObjectives for use must be set measurably––ImplementedAI Policy 5
A.9.4Intended use of the AI systemA.9YesHigh-risk systems only in line with the instructions––ImplementedAI Usage Policy 4; AI Usage Policy 10
A.10.2Allocating responsibilitiesA.10YesSeveral parties contribute to each system––ImplementedAI Policy 13; AI register
A.10.3SuppliersA.10YesThree systems are procured from suppliers––ImplementedAI Policy 14
A.10.4CustomersA.10YesCustomers interact with the chatbot––ImplementedAI Provider Policy 6; AI communication plan

Sources

Frequently asked questions

Do I have to implement all 38 controls?

ISO/IEC 42001 requires every control in Annex A to be considered. The necessary controls are included with a justification, the others excluded with a justification. An organisation that does not develop AI systems itself can, for example, exclude development controls, provided it justifies this.

Which reasons for exclusion are common?

The control concerns providers only, no system is high-risk, another control covers the risk, the risk is low, implementation is not feasible, or the control does not apply to the organisation. The reason alone is not enough; the justification explains it for the individual case.

What goes into the column "Reference"?

The section of the AI policy or the template in which the control is implemented. In an audit it leads directly to the evidence.

No Cookies!

This wiki collects nothing, bakes nothing and leaves nothing behind. There's nothing to consent to. Privacy doesn't get better than this.