The European Artificial Intelligence Act

The European Union’s AI Act establishes ethical guidelines and a regulatory framework for the development, deployment, and use of Artificial Intelligence (AI) systems in the European Union. In this whitepaper, we provide an overview of the key provisions outlined in the EU AI Act, addressing risk-classification, stakeholder considerations, and requirements, with a particular focus on safety-critical and therefore high-risk AI systems. In the scope of the EU AI Act, AI systems are categorized based on risk levels, ranging from minimal- to unacceptable- risk, each with its corresponding set of regulatory obligations. High-risk AI systems are subject to rigorous requirements spanning risk management, data governance, transparency, and human oversight. This whitepaper delves into the specifics of Articles 9 to 15 of the EU AI Act, covering the requirements for high-risk AI systems while also identifying gaps with respect to existing safety standards. We propose a framework for bridging these gaps by deriving concrete requirements from the EU AI Act inspired by contract-based design. We leverage our expertise in trustworthy AI and safety to develop a framework for deriving argumentation trees for generic properties of Machine Learning (ML) systems. We demonstrate this framework on three practical use cases across various sectors. Our work illustrates how AI systems in safety-critical domains such as automotive, industrial automation, and healthcare can meet regulatory standards while upholding ethical principles. The EU AI Act represents a significant step towards fostering trust, accountability, and responsible innovation in AI technologies. By following systematic verification processes for EU AI Act requirements, stakeholders can navigate the complex AI landscape with confidence, ensuring the ethical development and deployment of AI systems while safeguarding human interests and values.