AI stan­dard­i­s­a­tion

Standardisation plays a central role within the EU internal market. The EU’s new legislative framework is key to EU product regulation, defining the requirements for the safety and quality of products to enable the free movement of goods in the internal market. European legislation is restricted within this framework to laying down the essential requirements (protection objectives) without setting out technical details of how to achieve these requirements. The technical details are developed on the basis of standardisation requests from the European Commission by the stakeholders (from industry, associations, academia, society and the public sector) within the European standardisation organisations CEN, CENELEC and ETSI in accordance with the principle of consensus (see the WTO’s principles for the development of standards) and are then listed by the European Commission as harmonised European standards (hENs).

The division of work under the new legislative framework ensures that the protection objectives laid down in the legislation are implemented in a dynamic and innovation-friendly way and in line with the current state of the art. The application of harmonised European standards is generally voluntary for manufacturers; however, if a product meets the requirements in the standards, the presumption of conformity applies and it is assumed in cases of liability and vis-à-vis market surveillance authorities that the product meets the statutory requirements.

The Bundesnetzagentur has been continuously involved in standardisation activities since the beginning through its participation in relevant national, European and international bodies.

The European AI Act transposes the system under the new legislative framework to the field of AI. The risk-based approach under the framework is reflected in the AI Act in that the higher the risk of harm posed by a system to safety and fundamental rights, the higher the requirements to be met by the system. High-risk AI systems must therefore meet strict safety and quality requirements, with the technical details being set out in harmonised European standards. Standards therefore ensure that AI systems across the EU are developed and deployed in accordance with the same quality and safety requirements. This creates clarity for industry as well as trust among users. The standards lay down how to design AI systems safely in terms of, for example, data protection, transparency and clarity of decisions. Clear standards help companies to develop and market products more quickly as they know exactly which requirements they need to meet.

In this context, the focus of AI standardisation is on the ten fields defined in the European Commission’s standardisation request (see below). The aim of the standardisation request is to create harmonised standards that will play a key role in implementing the AI Act. This will strengthen the digital internal market since it will enable companies to offer their AI solutions in all EU countries under the same conditions.

Article 40 of the AI Act provides the framework for harmonised standards and other technical deliverables in the context of the AI Act. The article defines how conformity with harmonised standards can contribute to conformity with the requirements of the AI Act. AI systems that are in conformity with the harmonised standards are considered to be in conformity with the relevant technical requirements set out in the AI Act. This creates above all legal certainty for manufacturers.

Article 40 of the AI Act provides the framework for harmonised standards and other technical deliverables in the context of the AI Act. The article defines how conformity with harmonised standards can contribute to conformity with the requirements of the AI Act. AI systems that are in conformity with the harmonised standards are considered to be in conformity with the relevant technical requirements set out in the AI Act. This creates above all legal certainty for manufacturers.

Work on developing harmonised European standards for the AI Act began within the relevant European committee (CEN/CLC/JTC 21) following the first standardisation request for high-risk AI systems issued by the European Commission to CEN and CENELEC in May 2023. The German joint working committee for artificial intelligence (NA 043-01-42 GA) plays a particular role in this context. DIN, the German Institute for Standardization, and DKE, the German Commission for Electrical, Electronic and Information Technologies of DIN and VDE, work together with experts from various branches of industry within the national committee. They coordinate activities at national level and link them with those at European and international level, mirroring the work within the European committee CEN/CLC/JTC 21 and the international committee ISO/IEC JTC 1/SC 42.

Standardisation in the field of artificial intelligence needs to be driven forward at a swift pace. Technological, societal, legal and ethical aspects as set out in the AI Act need to be taken into account.
The Bundesnetzagentur will support this process and is open to dialogue with German players in the field of standardisation with the aim of balancing interests and developing proposals for a common understanding.