Cyber, data & AI

AI governance

AI governance is the set of policies, roles, responsibilities, decision-making and oversight through which an organisation directs and controls the development, procurement and use of AI systems, in line with its strategy, risk appetite and the law. It covers, among other things, the AI inventory, the classification of each system, the controls and accountability for them. ISO/IEC 42001:2023 sets requirements for an AI management system.

Source: ISO/IEC 42001:2023; Regulation (EU) 2024/1689 (AI Act) Articles 4, 9, 17, 26, 27 and 40; NIST AI RMF 1.0 (2023); IIA AI Auditing Framework (2024); Dutch Corporate Governance Code 2025 (explanatory notes to provision 1.2.1); EIOPA, Opinion on AI governance and Risk Management (EIOPA-BoS-25-360, 2025)

The EU AI Act (Regulation (EU) 2024/1689) does not prescribe a governance model, but it presupposes one. Providers and deployers take measures to support the AI literacy of their staff (Article 4). From 2 December 2027 (Annex III) or 2 August 2028 (Annex I), providers of high-risk systems establish a Risk Management system and a quality management system (Articles 9 and 17). Deployers assign human oversight to competent people and, in certain cases, carry out a fundamental rights impact assessment (Articles 26 and 27). ISO/IEC 42001:2023 provides the management system, with an AI risk assessment, risk treatment and an impact assessment for each system, and it is certifiable. The NIST AI Risk Management Framework 1.0 (2023) makes Govern a function that runs through all the others.

In the Netherlands, the explanatory notes to best practice provision 1.2.1 of the Dutch Corporate Governance Code 2025 cite the risks of the ethically responsible use of new technologies, such as Responsible AI, as an example of risks for which the management board sets the risk appetite; the supervisory board oversees this. For insurers, EIOPA (2025) sets out AI governance for systems that are not high-risk, in areas such as data governance, documentation, explainability and human oversight. In a consultation report (June 2026), the Financial Stability Board (FSB) asks the board and senior management to align AI adoption with the business model, risk appetite and strategy.

AI governance is broader than compliance with the AI Act: it also covers low-risk applications. The difference from data governance is that the latter concerns the data, whereas AI governance concerns the systems that use the data. Internal audit can assess AI governance using the IIA's AI Auditing Framework (updated in September 2024), which distinguishes governance, management and internal audit as domains. There is no Topical Requirement on AI.

Last updated: · Editor: ONE Risk Advisory

Frequently asked questions

Is an ISO/IEC 42001 certificate enough to comply with the AI Act?

No. ISO/IEC 42001 is an international standard for an AI management system, not a harmonised European standard. A harmonised standard only gives a presumption of conformity once it is referenced in the Official Journal (Article 40), and the European Commission's standardisation page (updated 3 August 2026) listed no AI standard with that status. A certificate shows that the management system meets the standard within the declared scope.

Who is responsible for AI governance in an organisation?

The management board sets policy and risk appetite; the supervisory board oversees (Dutch Corporate Governance Code 2025, provision 1.2.1). Each AI system should have an owner in the first line, with the second line providing advice, challenge and monitoring and internal audit providing independent assurance. For insurers, EIOPA (2025) states that the compliance and internal audit functions verify that the use of AI complies with applicable laws and regulations.

Is there a mandatory framework for auditing AI?

No. The IIA has no Topical Requirement on AI, although the Global Internal Audit Standards apply to every engagement. The IIA's AI Auditing Framework (2024) is practical guidance. Criteria often used are the AI Act, ISO/IEC 42001 and the NIST AI Risk Management Framework; the auditor records which version of which document served as a criterion.

← All terms