Appearance
ISO/IEC 42001:2023 — Annexes A–D
ISO/IEC 42001:2023 gives its requirements in Clauses 4 to 10, and it has four annexes after them. Annex A and Annex B are normative, and Annex C and Annex D are informative. This page gives the function of each annex. It also gives the use of Annex A and Annex B in the AI risk treatment of the organization (Clause 6.1.3).
Primary source
ISO/IEC 42001:2023, Information technology — Artificial intelligence — Management system, first edition 2023-12: Annexes A to D, and Clause 6.1.3 for the use of Annex A and Annex B. To read the annexes, get the standard from the ISO catalog page.
Normative and informative annexes
The Contents section of ISO/IEC 42001:2023 and each annex mark the four annexes as normative or informative:
| Annex | Name | Normative or informative | Use by the organization |
|---|---|---|---|
| A | Reference control objectives and controls | Normative | In its AI risk treatment, the organization must compare its necessary controls with Annex A (6.1.3 b)). It must also consider the Annex A controls that are "relevant" to implement the AI risk treatment options that it selected (6.1.3 c)). |
| B | Implementation guidance for AI controls | Normative | The organization must consider the guidance about how to implement the controls that it determined (6.1.3 e)). |
| C | Potential AI-related organizational objectives and risk sources | Informative | The organization can consider the objectives and risk sources in its risk management (C.1). |
| D | Use of the AI management system across domains or sectors | Informative | The organization can use the annex for its AI management system (AIMS) in a sector. It can also use the annex for the AIMS together with other management system standards (D.1, D.2). |
Annex A is normative, but this does not make each Annex A control mandatory. The organization determines the controls that are necessary for its AI risk treatment. A.1 tells the organization that it can also design and implement other controls (see Annex A). Annex A of ISO/IEC 27001:2022 is also normative. A different page compares ISO/IEC 42001 and ISO/IEC 27001.
Annex A — Reference control objectives and controls
Annex A of ISO/IEC 42001:2023 is normative. Table A.1 of Annex A has 38 controls in nine areas, A.2 to A.10. Each area has a set of controls with numbers. Each area has one control objective, but area A.6 has two, one for A.6.1 and one for A.6.2.
Clause A.1 gives the function of Table A.1 and its limits:
- The controls of Table A.1 are reference controls. They help the organization with its objectives, and with risk treatment for the design and operation of AI systems.
- The use of all control objectives and controls in Table A.1 is not mandatory. The organization can design and implement other controls (A.1 refers to 6.1.3).
- Annex B gives implementation guidance for each control in Table A.1.
Table A.1 gives each control as a requirement ("shall"). An Annex A control applies to the organization when its AI risk treatment selects the control as necessary (6.1.3).
| Area | Control | Name of the control |
|---|---|---|
| A.2 Policies related to AI | A.2.2 | AI policy |
| A.2 Policies related to AI | A.2.3 | Alignment with other organizational policies |
| A.2 Policies related to AI | A.2.4 | Review of the AI policy |
| A.3 Internal organization | A.3.2 | AI roles and responsibilities |
| A.3 Internal organization | A.3.3 | Reporting of concerns |
| A.4 Resources for AI systems | A.4.2 | Resource documentation |
| A.4 Resources for AI systems | A.4.3 | Data resources |
| A.4 Resources for AI systems | A.4.4 | Tooling resources |
| A.4 Resources for AI systems | A.4.5 | System and computing resources |
| A.4 Resources for AI systems | A.4.6 | Human resources |
| A.5 Assessing impacts of AI systems | A.5.2 | AI system impact assessment process |
| A.5 Assessing impacts of AI systems | A.5.3 | Documentation of AI system impact assessments |
| A.5 Assessing impacts of AI systems | A.5.4 | Assessing AI system impact on individuals or groups of individuals |
| A.5 Assessing impacts of AI systems | A.5.5 | Assessing societal impacts of AI systems |
| A.6.1 Management guidance for AI system development | A.6.1.2 | Objectives for responsible development of AI system |
| A.6.1 Management guidance for AI system development | A.6.1.3 | Processes for responsible AI system design and development |
| A.6.2 AI system life cycle | A.6.2.2 | AI system requirements and specification |
| A.6.2 AI system life cycle | A.6.2.3 | Documentation of AI system design and development |
| A.6.2 AI system life cycle | A.6.2.4 | AI system verification and validation |
| A.6.2 AI system life cycle | A.6.2.5 | AI system deployment |
| A.6.2 AI system life cycle | A.6.2.6 | AI system operation and monitoring |
| A.6.2 AI system life cycle | A.6.2.7 | AI system technical documentation |
| A.6.2 AI system life cycle | A.6.2.8 | AI system recording of event logs |
| A.7 Data for AI systems | A.7.2 | Data for development and enhancement of AI system |
| A.7 Data for AI systems | A.7.3 | Acquisition of data |
| A.7 Data for AI systems | A.7.4 | Quality of data for AI systems |
| A.7 Data for AI systems | A.7.5 | Data provenance |
| A.7 Data for AI systems | A.7.6 | Data preparation |
| A.8 Information for interested parties of AI systems | A.8.2 | System documentation and information for users |
| A.8 Information for interested parties of AI systems | A.8.3 | External reporting |
| A.8 Information for interested parties of AI systems | A.8.4 | Communication of incidents |
| A.8 Information for interested parties of AI systems | A.8.5 | Information for interested parties |
| A.9 Use of AI systems | A.9.2 | Processes for responsible use of AI systems |
| A.9 Use of AI systems | A.9.3 | Objectives for responsible use of AI system |
| A.9 Use of AI systems | A.9.4 | Intended use of the AI system |
| A.10 Third-party and customer relationships | A.10.2 | Allocating responsibilities |
| A.10 Third-party and customer relationships | A.10.3 | Suppliers |
| A.10 Third-party and customer relationships | A.10.4 | Customers |
A.10.4 Customers is part of Table A.1 in the first edition of 2023. In A.10.4, the organization must consider the needs and expectations of its customers in the development and use of its AI systems with responsibility. The guidance in B.10.4 applies when the organization supplies an AI-related product or service. The A.10.4 control has no such condition.
This table shows where some items are in ISO/IEC 42001:2023:
| Item | Position in ISO/IEC 42001:2023 |
|---|---|
| Data quality | A.7.4 Quality of data for AI systems. A.6 has no control for data quality. |
| Information for interested parties, external reporting, and communication of incidents | A.8.2 to A.8.5. In A.6.2.7, the organization must determine the necessary technical documentation for each group of interested parties that is "relevant", for example "users", partners, and supervisory authorities. It must give this documentation to them in the "appropriate" format. Area A.6 has no control for A.8.3 (external reporting of "adverse impacts" by interested parties) or for A.8.4 (communication of incidents). |
| System and performance monitoring | A.6.2.6 AI system operation and monitoring. A.9 has no control for monitoring. |
| Decommissioning | Decommissioning has no control. A.4.6 includes decommissioning in the work for which the organization documents its human resources. C.3.6 gives problems with decommissioning as an example of a life cycle risk source. |
| Human oversight | No control of Table A.1 has this name. The Annex B guidance for some controls includes human oversight (for example B.9.3). |
| Customer requirements in contracts | The B.10.4 guidance tells the organization that customer requirements can be part of contracts. The A.10.4 control does not refer to contracts. |
| Critical infrastructure | Table A.1 does not refer to critical infrastructure. |
How to use Annex A
In ISO/IEC 42001:2023, Annex A is a set of reference controls for the AI risk treatment (Clause 6.1.3). The organization must determine the controls that are necessary for its risk treatment options. Then it must compare them with Annex A, to make sure that no necessary control is missing (6.1.3 b)). The page Clauses 4 to 10 gives the full 6.1.3 process.
The organization must make a statement of applicability as an output of this AI risk treatment process (6.1.3 f)). The page Scope and certification gives the contents of the statement of applicability.
Annex B — Implementation guidance for AI controls
Annex B of ISO/IEC 42001:2023 is normative. Annex B uses the areas of Table A.1. For each area, Annex B gives the control objective. For each control, it gives the control and implementation guidance, and for some controls, other information. Table A.1 gives each control with "shall", and Annex B gives the same control with "should".
Clause B.1 gives these rules for the guidance:
- The guidance gives information about how to implement the controls of Table A.1, and about the control objective.
- It is not necessary for the organization to document or justify the inclusion or exclusion of the guidance in the statement of applicability.
- The guidance is not always applicable or sufficient in all conditions. It does not always give all that is necessary for the specified control requirements of the organization.
- The organization can extend or change the guidance. It can also use a different method to implement a control for its requirements and its risk treatment.
- The organization can determine more controls than the controls in Annex B (B.1 refers to 6.1.3).
- The organization can use Annex B as the first step when it implements controls.
How to use Annex B
In its AI risk treatment, the organization must consider the Annex B guidance when it implements controls (6.1.3 e)). This duty applies to the controls that the organization determined in 6.1.3 b) and c). It is a duty to consider the guidance. The organization can use the guidance without change, change it, or use a different method to implement the control (B.1). The page Clauses 4 to 10 gives the other steps of 6.1.3.
Annex C — Potential AI-related organizational objectives and risk sources
Annex C of ISO/IEC 42001:2023 is informative. Annex C gives possible organizational objectives (C.2) and risk sources (C.3), each with a short text. The organization can consider these items in its risk management (C.1). C.1 also gives these limits:
- The sets of objectives and risk sources are not complete, and not all items apply to all organizations.
- ISO/IEC 42001 recommends that the organization determines which objectives and risk sources apply to it.
- ISO/IEC 23894 gives more information about the objectives, the risk sources, and their relation to risk management.
C.2 gives 11 objectives:
- C.2.1 Accountability
- C.2.2 AI expertise
- C.2.3 Availability and quality of training and test data
- C.2.4 Environmental impact
- C.2.5 Fairness
- C.2.6 Maintainability
- C.2.7 Privacy
- C.2.8 Robustness
- C.2.9 Safety
- C.2.10 Security
- C.2.11 Transparency and explainability
C.3 gives seven risk sources:
- C.3.1 Complexity of environment
- C.3.2 Lack of transparency and explainability
- C.3.3 Level of automation
- C.3.4 Risk sources related to machine learning
- C.3.5 System hardware issues
- C.3.6 System life cycle issues
- C.3.7 Technology readiness
How to use Annex C
ISO/IEC 42001:2023 refers to Annex C in these clauses:
- The NOTE to 6.2 refers to Annex C for a set of AI objectives that are related to risk management. That set is not complete. The page Clauses 4 to 10 gives the 6.2 duties for AI objectives.
- The part "Other information" of B.6.1.2 and of B.9.3 refers to Annex C. Annex C can help the organization to determine its objectives for AI development and for AI use.
- D.1 refers to Annex C for possible objectives for the development and use of AI with responsibility.
Clause 6.1.2 (AI risk assessment) and Clause 6.1.4 (AI system impact assessment) do not refer to Annex C. Annex C does not refer to Clause 6.1.4 or to A.5. Clause 6.1.4, A.5.4, and A.5.5 give the impacts on individuals, groups of individuals, and societies.
Annex D — Use of the AI management system across domains or sectors
Annex D of ISO/IEC 42001:2023 is informative. It has two parts, D.1 and D.2.
D.1 gives these points:
- The AI management system (AIMS) can apply to each organization that makes, supplies, or uses products or services with an AI system. D.1 gives these examples of sectors: "health", "defence", "transport", "finance", "employment", and "energy".
- ISO/IEC 42001 gives the AI-specific part. Generic or sector-specific management system standards usually consider an objective for all types of system, and not for AI only.
- An AI system can contain components that do not use AI. Thus, the development and use of an AI system with responsibility includes the full system, and not only the AI part. Information security applies to AI in general, because AI processes information.
- ISO/IEC 42001 recommends that the organization works on safety, security, privacy, environmental impact, and other such objectives together for the full system. It recommends that the organization does not work on them independently for the AI part and for the other parts.
- D.1 tells the organization that integration with generic or sector-specific management system standards is necessary for the development and use of an AI system with responsibility, for the "relevant topics". Annex D is informative. D.1 is guidance and sets no requirement.
D.2 gives other management system standards that the organization can use with the AIMS:
| Standard | Information in D.2 |
|---|---|
| ISO/IEC 27001 (information security) | An organization with an AIMS can want to work on security objectives with the same method. Then it can implement ISO/IEC 27001 for its information security management system. The two standards use the harmonized structure (the name in D.2 is high-level structure). D.2 tells the organization that this helps it to use the two standards together, and that their use together has much value for the organization. When the organization implements the controls that are related in part to information security (see B.6.1.2), it can do this together with its ISO/IEC 27001 work. |
| ISO/IEC 27701 (privacy) | Integration with ISO/IEC 27701 can have value for the organization. The organization can make the privacy objectives and privacy controls of the AIMS (B.2.3, B.5.4) part of its ISO/IEC 27701 work. |
| ISO 9001 (quality) | The AIMS can add to the ISO 9001 requirements, for example on risk management, software development, and the supply chain. |
| ISO 22000 (sector example) | For an AI system that is used to make, prepare, and move "food", ISO 22000 and an AIMS are each "relevant". |
| ISO 13485 and IEC 62304 (sector examples) | The AIMS can help with the requirements for medical device software in ISO 13485. It can also help with the requirements of other standards of the medical sector, for example IEC 62304. |
NOTE 5 to Clause 6.1.1 also refers to the sectors of Annex D. It tells the organization that risk (see 3.7) has a wide scope and can apply to all sectors. Annex D does not give guidance about how to divide an AIMS between parts of the organization or types of AI system. That decision is for the organization, in the scope of its AIMS (Clause 4.3).
The Requirements in Modulos
Modulos has two Frameworks for ISO/IEC 42001. The two Frameworks have the name ISO 42001: OFF-10 at the Organization level and MFF-10 at the AI System level. The table on Operationalizing ISO/IEC 42001 in Modulos gives the Requirement for each clause and each Annex A control.
Where to go next
ISO/IEC 42001
The standard, the organizations that it applies to, its structure, certification, and how it compares with the EU AI Act
Scope and certification
The AI roles (Clause 4.1), the AIMS scope (Clause 4.3), the statement of applicability, and certification bodies
Clauses 4 to 10
The requirements of Clauses 4 to 10, with the AI risk assessment, the AI risk treatment, and the AI system impact assessment
Operationalizing in Modulos
OFF-10 and MFF-10, the set-up decisions, the Requirement for each clause and Annex A control, and the steps
Sources of this page
- ISO/IEC 42001:2023, Information technology — Artificial intelligence — Management system, first edition 2023-12 (ISO catalog page). This page uses these parts of it:
- The Contents section.
- Clauses 3.7, 3.26, 4.3, 6.1.1 (NOTE 5), 6.1.2, 6.1.3, 6.1.4, and 6.2.
- Annex A: A.1 and Table A.1.
- Annex B: B.1, B.2.3, B.5.4, B.6.1.2, B.9.3, and B.10.4.
- Annex C: C.1 to C.3.
- Annex D: D.1 and D.2.
- ISO/IEC 27001:2022, edition 3 (2022-10): Clause 6.1.3 and Annex A.
Disclaimer
This page gives general information only. It is not legal advice. It does not replace the audit of a certification body. Compare this page with the text of ISO/IEC 42001:2023.