November 25, 2025
TÜV SÜD now offers voluntary conformity certificates for AI systems/products. This service is aimed at manufacturers who wish to obtain an independent assessment of their technical documentation and product conformity in accordance with the EU AI Act (EU Regulation 2024/1689) ahead of the mandatory conformity procedure. Additional services such as training and gap assessments, help to ensure efficient compliance with the EU AI Act requirements.The voluntary conformity certificate confirms that the tested product meets some or all of the general technical requirements of the EU AI Regulation. Additionally, the specific requirements necessary to support the technical documentation in accordance with Article 11 of the EU AI Act are reviewed. This new service offers manufacturers a valuable opportunity to evaluate their AI products early on with regard to the EU AI Act requirements, thereby supporting them in efficiently preparing for the legally required conformity procedures.
Step by step towards compliance with the EU AI Act
The certificate does not replace the legally required EU Declaration of Conformity (DoC) and does not entitle the manufacturer to affix the CE mark. However, manufacturers receive valuable information on the legal compliance maturity of their AI system. Only once the technical documentation has been finalised and conformity with all applicable directives has been confirmed can the manufacturer issue an EU Declaration of Conformity and affix the CE marking under their own responsibilty. AI products/applications in higher risk classes must have their conformity verified by a notified body. However, the procedures for establishing and designating these notified bodies are still ongoing.Further information about the Attestation of Conformity (AoC) service and related services is available from [email protected].
EU AI Act – Deadlines are approaching
The EU AI Act will come into force in stages, depending on the risk classification of an AI system. Here are the most important dates by risk class:
Risk category | What applies? | Effective date |
Unacceptable risk | The use of these AI systems (e.g., social scoring by governments, certain types of biometric surveillance) is prohibited. | February 2, 2025 |
General-Purpose AI | Certain transparency requirements and codes of practice for general-purpose AI (GPAI) will gradually take effect. | From August 2, 2025 |
High risk | Strict requirements (e.g., risk assessment, documentation requirements, ongoing review) for AI in education, medicine, government agencies, etc. | December 2, 2027 (Annex III)* |
High risk (products as safety components) | Particularly affects AI as part of regulated products (toys, medical devices, etc.). | August 2, 2028* |
Limited/minimal risk | Hardly any/no requirements, mostly transparency obligations, such as information on AI interaction. | Immediately or with general launch in 2026 |
*These dates represent the final deadlines for high-risk AI obligations; however, according to the Digital Omnibus Package released on November 19th, they may be brought forward if conformity assessment tools are approved earlier.
The AI Act Risk Navigator, a free online tool from TÜV SÜD and TÜV AI.Lab, provides initial guidance on the risk classification of AI systems and models: tuv-sud.risk-navigator.ai.
More information: tuvsud.com/en/topics/artificial-intelligence
Press contact: Dirk Moser-Delarami