Auditing of AI in Railway Technology – a European Legal Approach

Abstract

Artificial intelligence (AI) promises major gains in productivity, safety and convenience through automation. Despite the associated euphoria, care needs to be taken to ensure that no immature, unsafe products enter the market, especially in high-risk areas. Artificial intelligence systems are therefore to be integrated into the European Union’s existing product safety system by the planned AI regulation. This is accomplished by horizontally linking the draft AI regulation (AI-Act) with the existing harmonizing legal acts that ensure the safety of products and provide for a complex system of approval and testing concepts for this purpose. The railway sector is no exception, which is why potential AI systems for monitoring tracks or simple and accurate train detection are also subject to this approval regime. The following article highlights the challenges that exist in the railway sector on the path to verifiable AI systems in this regulatory context.