Published on February 4, the Guidelines on the prohibited systems under the Artificial Intelligence (AI) Act aim to provide legal certainty, promote the consistent, effective, and uniform application of the regulation, and ensure compliance with its obligations.
The Commission has published new guidelines to facilitate the understanding of the prohibited practices defined in Article 5 of Regulation (EU) 2024/1689 on Artificial Intelligence.
The AI Act follows a risk-based approach, classifying AI systems into four different risk categories. Among these, there are prohibited AI systems, which entail unacceptable risks to fundamental rights and the values of the Union. The prohibited practices include, among others, the use of AI-enabled manipulative techniques, social scoring systems by public or private actors that may result in discriminatory outcomes, emotion recognition systems in situations related to the workplace and education, and real-time remote biometric identification systems, except in exhaustively listed situations.
These guidelines, published on February 4, 2025, address the prohibited AI systems, whose provisions are fully applicable since February 2, 2025. Even if the provisions on penalties for non-compliance will not become applicable until next August, Chapters I (general provisions) and II (prohibited practices) of the AI Act are already mandatory for providers and deployers of AI systems. Therefore, companies developing or using AI systems in the course of their activities must adopt the necessary measures to ensure they do not place on the market, put into service or use prohibited AI systems.
The goal is to provide legal certainty, promote the consistent, effective, and uniform application of the regulation, and ensure compliance with its obligations. For each of the prohibited practices defined in Article 5, the guidelines offer legal explanations of the main components of the provision, practical examples, and clarifications on practices that fall outside the scope of the prohibitions, as well as suggesting potential measures to avoid their use.
Additionally, the guidelines identify areas where the prohibitions overlap with other legal acts of the European Union. They also address general exceptions, such as those related to systems used exclusively for military, defense, or national security purposes, which are relevant for the practical application of these categories.
Although these guidelines are non-binding, they provide valuable insights into the Commission’s interpretation of the prohibitions and serve as practical guidance for developers, deployers, and various operators in the AI value chain, who must assess their AI systems and comply with the obligations under the regulation. Furthermore, the guidelines will assist market surveillance authorities in their control and enforcements functions.
With these guidelines, the Commission takes a further step in the practical implementation of the AI Act, which is now a reality.