European AI Standards - Technical Standardization and Implementation Challenges under the EU AI Act

Abstract

Harmonized standards are the cornerstone of efficient EU AI Act compliance. This paper presents one of the first systematic analyses of European technical and soon to be harmonized standardization for organizations providing AI systems. Based on in-depth qualitative interviews with 23 leading European organisations developing AI applications across different sectors, such as Mistral and Helsing, and providing transparency regarding the status quo of draft standards, it examines how companies, especially startups and SMEs, are dealing with the contemplated standardization under the EU AI Act and sectoral standardization following vertical product safety legislation. Industry sectors covered include mobility, finance, manufacturing, healthcare, enterprise software, as well as defense and legal tech. Key challenges identified comprise an effective implementation period of only approximately 6 months compared to at least 12 months actually required, an imbalance of participation and influence in standardization committees, double regulation and technical implementation hurdles as well as significant annual costs for harmonized standards compliance. Technical standards are currently reshaping global AI competition and will have a massive influence on the AI landscape as market entry barriers, particularly on startups. Hence, the paper offers concrete policy recommendations based on the revealed challenges for AI providers.

Keywords: AI Standardization, European AI Standards, EU AI Act, AI Act, Technical Standards, Harmonized Standards, AI Compliance, Digital Regulation, AI Trustworthiness, European AI Policy, CEN-CENELEC JTC 21, High-Risk AI Systems