EU AI Act and ISO 42001: how they connect

The EU AI Act sets requirements for AI systems. ISO 42001 provides the framework to meet them. Learn how they connect and what your organization needs to do.

EU AI Act and ISO 42001: how they connect

The EU AI Act (AI Act) entered into force on 1 August 2024 and is the EU’s main law on artificial intelligence. In parallel, ISO published ISO/IEC 42001 for AI management systems. Together they form a pair: the law says what you must do, the standard shows how.

This article explains what the AI Act requires, what ISO 42001 contains, and how to use the standard to meet the law’s requirements.

What is the EU AI Act?

The AI Act (Regulation (EU) 2024/1689) applies directly in all EU member states. The law covers everyone who develops, provides, or uses AI systems within the EU.

The core of the law is a risk-based classification in four tiers:

  • Unacceptable risk – AI systems that manipulate people, exploit vulnerabilities, or perform social scoring. The ban has been in effect since 2 February 2025.
  • High risk – AI systems in areas such as recruitment, credit scoring, education, and critical infrastructure. These systems must meet requirements for risk management, documentation, transparency, and human oversight. The requirements apply from 2 August 2026.
  • Limited risk – AI systems that interact directly with people, such as chatbots. Transparency requirements apply, meaning users must be informed they are communicating with an AI.
  • Minimal risk – All other AI systems. No specific requirements, but the law encourages voluntary codes of conduct.

The ban exists because these AI systems harm human autonomy, exploit vulnerabilities, or undermine democratic institutions. Violations can result in fines of up to EUR 35 million or 7% of global annual turnover.

Timeline

Date What applies
1 August 2024 Law enters into force
2 February 2025 Ban on AI systems with unacceptable risk
2 August 2025 Requirements for general-purpose AI models (GPAI)
2 August 2026 Main body of the law, including high-risk AI requirements (see note below)
2 August 2027 High-risk classification via EU harmonised product legislation

November 2025 update, Digital Omnibus: On 19 November 2025, the European Commission proposed (COM(2025) 868) that the date for high-risk rules could be extended by up to 16 months beyond 2 August 2026, provided the required standards and support tools are in place. This means the rules would apply at the latest in December 2027, but potentially earlier. The proposal has not yet been adopted by Parliament and Council. The other dates in the timeline above are unaffected.

What is ISO 42001?

ISO/IEC 42001:2023 is a certifiable standard for AI management systems (AIMS). It follows the same structure as other ISO management system standards (ISO 9001, ISO 27001, ISO 14001) with clauses 4-10.

For organisations using AI, assessment starts with the task. A tool summarising meeting notes needs a different assessment from the same tool screening job applications. The risk assessment in 6.1.2 helps you prioritise what could go wrong. Risk treatment under 6.1.3 concerns how you address those risks. One possible choice is for a recruiter to compare an AI recommendation with the original application before anyone is excluded.

The impact assessment in 6.1.4 adds the perspective of those affected. Faster recruitment may benefit the company while disadvantaging candidates. Investigate who could be wrongly excluded and how that could be detected and corrected. These examples are application advice, not a complete account of the requirements.

Annex A supports the choice of measures. For a purchased tool, you might need to clarify who approves new uses, what information colleagues need and what the supplier is responsible for. The statement of applicability, SoA, makes your choices and reasons traceable; it does not replace assessment of actual use.

How ISO 42001 supports AI Act compliance

ISO 42001 can help you coordinate the work behind legal obligations. A risk discovered during testing needs an owner, a response and a check that the response helped. This lets you trace how a decision was made rather than just show a policy.

Providers of high-risk AI need to address risk management, data quality, technical documentation, information for users and human oversight under AI Act Articles 9–14, as well as quality management under Article 17. Your own obligations depend on your role and the system’s use. ISO 42001 risk assessment and treatment (6.1.2–6.1.3) can support the work, but an assessment under the standard does not by itself show compliance with the law.

ISO 42001 certification is not a legal requirement. But it gives you a demonstrable structure showing that you work systematically with AI governance. During a regulatory inspection, it can be the difference between showing a functioning system and having to start building one.

Where to start

If your organization uses AI systems, including purchased ones, start with these steps:

  1. Take inventory: which AI systems do you use? Many organizations have more AI systems than they realize, from chatbots and analytics tools to AI features embedded in existing software.

  2. Classify according to the AI Act: assess which risk category each system falls into. Systems in recruitment, credit scoring, education, and similar areas likely fall into the high-risk category.

  3. Start the risk assessment: use ISO 42001’s framework (clause 6.1.2) to assess risks and consequences. Go through the Annex A controls and determine which apply.

  4. Document your assessment: record why you chose a tool and which limitations you found. Check documentation needs against the requirements applicable to your use.

  5. Decide whom colleagues should contact when an AI response is wrong or a new use is proposed.

If you already have a management system, such as ISO 9001 or ISO 27001, you have a head start. ISO 42001 follows the same underlying structure, and much of the work on context, leadership, planning, and improvement overlaps.

GDPR applies in parallel

The Swedish Authority for Privacy Protection (IMY) points out that GDPR applies when personal data is processed in connection with the development or use of AI systems. The AI Act does not replace the data protection regulation. If your AI system processes personal data, you need to comply with both frameworks.

Summary

The EU AI Act sets the rules. ISO 42001 gives you the method to follow them. Start with an inventory of your AI systems, classify them, and perform a risk assessment. If you already work with ISO standards, you already have the foundation.

AmpliFlow participated in ISO’s working group for ISO/IEC 42001 and has built support for the standard, from risk analysis and statement of applicability to AI features that help you process information. Read more about how AmpliFlow supports ISO 42001 or book a demo.

Share this article

LinkedIn

This article is also available in Swedish.

Related articles

ISO 9001:2026: what changes for certified companies?

ISO 9001:2026: what changes for certified companies?

CRA: new EU requirements for software and connected products

CRA: new EU requirements for software and connected products

Eight AmpliFlow updates from summer 2026

Eight AmpliFlow updates from summer 2026