Targeted Stakeholder Consultation on Classification of AI Systems as High-Risk
This consultation is aimed at gathering insights from various stakeholders regarding the classification of AI systems as high-risk, a significant aspect of the recently enacted Artificial Intelligence Act (the ‘AI Act’), which became effective on August 1, 2024.
Purpose of the Consultation
The primary goal of this consultation is to collect input from stakeholders, which includes providers and deployers of high-risk AI systems, industry organizations, academia, independent experts, civil society organizations, and public authorities. The feedback will help clarify practical examples and issues related to the guidelines for classifying high-risk AI systems and future requirements.
Overview of the AI Act
The AI Act establishes a single market and harmonized rules for trustworthy and human-centric AI within the European Union. Its objectives are to promote innovation and the adoption of AI technologies while ensuring a high level of protection for health, safety, and fundamental rights, including democracy and the rule of law.
One of the key features of the AI Act is its risk-based approach, which categorizes AI systems into different risk levels, including high-risk systems as specified in Chapter III. The obligations for these high-risk systems will come into effect two years after the act’s implementation, specifically from August 2, 2026.
Classification of High-Risk AI Systems
The AI Act distinguishes between two categories of AI systems considered high-risk:
- Article 6(1): AI systems embedded as safety components in products or that are products covered by Union legislation in Annex I, which may adversely impact health and safety.
- Article 6(2): AI systems that pose significant risks to health, safety, or fundamental rights based on their intended purpose.
Annex III of the AI Act outlines eight areas where AI systems could potentially pose significant risks, along with specific use cases classified as high-risk. Article 6(3) provides exemptions for certain AI systems intended for specific cases listed in Annex III but do not pose significant risks.
Requirements for High-Risk AI Systems
AI systems classified as high-risk must adhere to several requirements detailed in Chapter III, Section 2, including:
- Data governance
- Documentation and record keeping
- Transparency and user information
- Human oversight
- Robustness, accuracy, and security
Providers of high-risk AI systems are obligated to ensure compliance with these requirements, which include implementing a quality management system and conducting a conformity assessment before market placement or service initiation. Additionally, deployers of these systems must meet obligations pertaining to correct usage, human oversight, and monitoring.
Guidelines and Future Directions
According to Article 6(5) of the AI Act, the Commission is tasked with providing guidelines for the practical implementation of high-risk classification rules by February 2, 2026. These guidelines will include a comprehensive list of practical examples of both high-risk and non-high-risk AI systems.
Furthermore, Article 96(1)(a) mandates the development of guidelines on the practical application of requirements for high-risk AI systems and operator responsibilities, including those along the AI value chain as per Article 25.
Structure of the Consultation Questionnaire
The consultation questionnaire is divided into five sections:
- Section 1: Classification rules of high-risk AI systems as per Article 6(1) and Annex I.
- Section 2: Classification of high-risk AI systems under Article 6(2) and Annex III.
- Section 3: General questions regarding high-risk classification.
- Section 4: Requirements and obligations for high-risk AI systems.
- Section 5: Need for amendments to the lists of high-risk use cases and prohibited AI practices.
Respondents are encouraged to provide explanations and practical cases to enhance the guidelines’ usefulness. The consultation will be open for 6 weeks, starting from June 6 to July 18, 2025.
Conclusion
All contributions to this consultation will be made publicly available, and individuals can request anonymity for their submissions. The results will be summarized and based on aggregated data, ensuring that respondents are not directly quoted.
