The adoption of the AI Act marks a major step forward in regulating artificial intelligence in Europe. It sets out important requirements for AI systems, including risk management, transparency, robustness, and human oversight. However, while these principles define what needs to be achieved, they do not always explain how organisations should implement them in practice.
This is where standards play a key role.
Standards act as a practical bridge between regulation and implementation. They translate high-level legal requirements into concrete technical specifications, processes, and methodologies that organisations can follow when developing and deploying AI systems. In this way, they provide clarity and support companies in turning regulatory obligations into actionable steps.
An important feature of this system is the concept of presumption of conformity. When organisations apply harmonised European standards, they can be considered compliant with certain requirements of the AI Act without needing to prove this from scratch. This reduces uncertainty and helps streamline compliance efforts.
Standardisation activities are currently being led by CEN-CENELEC, which has organised the work into thematic clusters such as quality management, risk management, trustworthiness, data, and cybersecurity. This structured approach reflects the interconnected nature of AI systems and their regulatory requirements.
At the same time, the development of these standards is not without challenges. The process is complex and still evolving. Among the key issues are the limited availability of specialised experts, the lack of widely agreed benchmarks and evaluation metrics, and differences between how AI systems are described in regulation and how they are actually developed and used in practice.
The CERTAIN project contributes to this ongoing effort by supporting the development and alignment of standards. Drawing on the expertise of partners actively involved in standardisation, the project helps bridge the gap between regulatory objectives and practical implementation, promoting a more consistent and effective application of the AI Act.
While progress may be slower and more complex than initially expected, these efforts are essential to ensure that the AI Act can be successfully applied in real-world scenarios and that AI systems in Europe are both innovative and trustworthy.