The Artificial Intelligence Act (AI Act) - a BusinessEurope position paper

Key messages

• We welcome the Commission’s risk-based approach outlined in the AI Act and share the ambition to ensure that AI is safe, lawful and in line with EU fundamental rights.
• We recommend carefully addressing all potentially unwanted consequences and administrative burden for industry which could discourage investment in the development of AI systems and hurt EU competitiveness in the long term.
• The proposed definition of “AI systems” is too broad. We recommend using the definition proposed by the High-Level Expert Group on AI, focusing on AI systems that display intelligent behaviour and take actions with some degree of autonomy.
• The proposed classification rules for high-risk AI should be refined to ensure consistency with sectoral legislation in Annex II. The AIA should only regulate high-risk AI applications in areas where a clear regulatory gap has been demonstrated.
• We agree that some AI use-cases in the areas listed in Annex III will require to undergo specific requirements, but we recommend adjusting the scope of application based on objective and clear criteria and to have a differentiation between applications of AI according to the actual risk they pose to fundamental rights and/or health and safety.
• Responsibilities of different actors in the AI value chain must be reassessed and clarified to ensure obligations are allocated to the actors that can ensure compliance.
• We support measures that can promote trustworthy AI in Chapter 2. However, we recommend reviewing the compliance framework to ensure 1) a proportionate risk-burden balance, 2) clear allocation of responsibilities, and 3) sufficient flexibility to adapt to new knowledge and evidence.
• We welcome the possibility to conduct internal-control checks and the use of consensus-based harmonised standards to demonstrate compliance. We recommend aligning with existing and ongoing international standardisation activities and certifications issued under a EU cybersecurity scheme.
• We recommend safeguarding the experimental nature of sandboxes schemes and converting their voluntary nature into an obligation, with well-established criteria to ensure an effective access to businesses, particularly SMEs.