Verständnis der Standardisierung des KI-Gesetzes

European harmonised standards translate legal requirements into technical specifications. They are developed by the European Standardisation Organisations (CEN-CENELEC and ETSI) following a request by the European Commission and can be used by providers or manufacturers of products intended for the EU market.

The harmonised standards for the AI Act will help AI providers to demonstrate compliance with legal requirements. Currently, harmonised standards are being developed for high-risk AI systems requirements, such as risk management, data quality, or human oversight.

European harmonised standards are developed by recognised European Standardisation Organisations: CEN, CENELEC, and ETSI.

In June 2021, CEN and CENELEC established a joint technical committee called JTC 21, dedicated to drafting AI standards, including harmonised standards for the AI Act. JTC 21 brings together experts from industry, society, and academia. Experts can join the standardisation work by applying through their National Standardisation Bodies. 

You can keep up to date by subscribing to the JTC 21 newsletter.

A product or system is presumed to satisfy the essential requirements of the relevant EU legislation, if it complies with harmonised standards referenced in the Official Journal of the European Union (OJEU). This ‘presumption of conformity’ shifts the regulatory burden of proof on public authorities to disprove conformity.  

Once harmonised standards for the AI Act are finalised and approved by the Commission, they will provide a clear path to demonstrate compliance for high-risk AI systems providers and reduce the risk of legal challenges. 

Harmonised standards, especially those which are conferred ‘presumption of conformity’, offer legal certainty under the AI Act, and support AI providers to develop compliant AI systems. In doing so, these standards support innovation and the development of trustworthy AI systems used in high-risk use cases. They foster trust and market acceptance, enabling developers who adopt them to be competitive while ensuring their solutions meet the highest safety standards. 

The European Commission has requested CEN and CENELEC to draft harmonised standards for the following areas:

• Risk management systems for AI systems 
• Governance and quality of datasets of AI systems 
• Record keeping through logging capabilities of AI systems 
• Transparency and information provisions for users of AI systems 
• Human oversight of AI systems 
• Accuracy specifications for AI systems 
• Robustness specifications for AI systems 
• Cybersecurity specifications for AI systems 
• Quality management systems for providers of AI systems, including post-market monitoring processes 
• Conformity assessment for AI systems 

If referenced in the Official Journal of the European Union, harmonised standards will be accompanied by an annex which maps the legal requirements in the AI Act to the clauses in the standard. 

The Commission is actively involved in monitoring the standardisation process, from drafting to publication.

All harmonised standards begin with a ‘standardisation request’ from the Commission to the standardisation bodies. The standardisation request sets the process in motion and ensures alignment with the legal text by clarifying which legal requirements the harmonised standards should aim to cover. 

While the Commission does not draft the harmonised standards, representatives are present as counsellors and observers to ensure the deliverables are aligned with the objectives of the AI Act and identify quality issues early-on. Several EU officials have been registered as ‘European Counsellors’ in JTC 21.

These representatives do not take part in the drafting of harmonised standards for the AI Act, but they do participate in plenary, working group, and task group meetings. They ensure that the harmonised standards and standardisation deliverables are aligned with the objectives of the AI Act.

After harmonised standards for the AI Act are published, the Commission will review them substantively to assess their compatibility with the requirements of the AI Act. If they are deemed to accurately reflect the legal provisions, the Commission will submit the harmonised standards for reference in the Official Journal of the European Union (OJEU), giving them presumption of conformity.

The first harmonised standards are expected to be published by CEN and CENELEC in 2026. After that, the Commission will start the review to assess whether the references to these standards can be published in the Official Journal of the European Union.

CEN and CENELEC are responsible for the planning of the harmonised standards. Find more information about the JTC 21 work on European AI standardisation.

The Commission has endorsed a General-Purpose AI Code of Practice and is facilitating the drawing-up of a Code of Practice on transparent AI systems. These codes of practice are voluntary tools designed to help industry comply with the AI Act’s obligations for general-purpose AI models (GPAI) and the transparency requirements of certain AI systems.  

Harmonised standards are similar to the codes of practice: they are practical tools to guide relevant stakeholders in complying with the AI Act. However, standards being currently developed focus specifically on high-risk AI systems and have the added benefit of presumption of conformity, which is unique to harmonised standards.  

There are several agreements between the European Standardisation Organisations and international standardisation organisations, such as the Vienna agreements between CEN and ISO. Through these agreements, European standards can reference international standards or adopt international standards in full as European Standards.  

Besides, European Standardisation experts are actively involved in building a broader global framework of AI standards by contributing to the work of various international standardisation organisations such as ISO, IEC or ITU. The Commission also provides training and funding to support experts participating in international standard development to align standardisation contributions with EU priorities. 

European standardisation efforts rely preferably on international standards. However, if international standards do not align with the legal requirements of the AI Act, new standards will provide guidance on compliance.  

For example, although ISO/IEC 42001:2023 helps to set up an AI management system, its goals and definitions are not aligned with the quality management system that is required under the AI Act. This is why the Commission has requested the development of a new standard for a quality management system that focuses on regulatory compliance with the AI Act.   

Yes. StandICT.eu is the recognised “go-to” initiative on ICT standards in Europe, directly supporting experts, strengthening Europe’s role in international Standard Developing Organisations (SDOs), and aligning standardisation contributions with EU priorities. Through its Fellowship Programme, StandICT.eu has channelled over €7 million to more than 900 European standards experts, enabling them to take leadership roles in global SDOs and make strategic contributions.  

The next StandICT call will focus on AI and harmonised standards. Find out more about StandICT.eu academy, fellowship opportunities and open calls.  

The Commission also provides funding to the European Standardisation Organisations in the form of EU action grants. These grants usually support targeted actions such as the development of specific harmonised standards, secretariat support, the development of communications activities or the subcontracting of experts to account for possible gaps in technical expertise or the inclusiveness of the standardisation process. Several grants have been attributed to CEN-CENELEC for the development of harmonised standards for the AI Act. 

Keep up to date on the latest grant opportunities via the EU Funding & Tenders Portal