Human-centred robotics and the EU AI Act: selected standards and implications

Abstract

Robotics and AI are key factors in enhancing business and national resilience, particularly in maintaining high-wage manufacturing in countries facing demographic challenges. Both are instrumental in making manufacturing more agile and flexible. Robots and humans will have to collaborate closely. Creating the necessary intelligent autonomous systems will go well beyond the typical considerations of machine learning. In line with the EU’s Industry 5.0 vision for a sustainable, human-centred and resilient European industry, humans and robots are expected to work so closely together that robot software must be designed with humans in mind from the outset. Thus, ethical, legal, and social implications (ELSI) must also be considered. Legal implications are multifaceted, ranging from AI Law, Product Liability Law, Product Safety Law, Machinery Law, to Technical Standards, Data Protection Law, Copyright and IP Law, Labour Law, etc. This contribution will briefly introduce the required technical features and, based on that, explore selected relevant legal implications and related standards of the new EU AI Act. The Act aims to promote human-centricity and entered into force on 1 August 2024, with its applicability phased in over a period of three years until 2 August 2027. Special references will be made to human-centredness, the subsumption of the plant owner to the categories of obligated parties (provider, product manufacturer, deployer, authorised representative or distributor), as well as the classification to the AI risk scheme (prohibited risks, high risks, transparency risks, general-purpose risks, and systemic risks as well as minimal risks) and related obligations. The high relevance of the AI Act for industrial human-robot settings will be shown. As AI evolves exponentially, so does its significance for industrial and national agility and resilience. Obligations vary widely according to risk classification and the obligated operator. Coordination and information flows between providers and deployers of related AI systems are key. As plant owners might become providers themselves, the duties could be more varied than initially assumed. Design and implementation of human-centred robotic scenarios, thus, should be well planned, structured, documented, and constantly evaluated. The respective AI models and AI systems need to be legally compliant and human-centred by design. An unprecedented dialogue across disciplinary fields is required.