Methods and Tools for the Engineering and Assurance of Safe Autonomous Systems (Dagstuhl Seminar 24151)

Autonomous systems rely increasingly on Artificial Intelligence (AI) and Machine Learning (ML) for implementing safety-critical functions. It is widely accepted that the use of AI/ML is disruptive for safety engineering methods and practices. Hence, the problem of safe AI for autonomous systems has received a significant amount of research and industrial attention over the last few years. Over the past decade, multiple approaches and divergent philosophies have appeared in the safety and ML communities. However, real-world events have clearly demonstrated that the safety assurance problem cannot be resolved solely by improving the performance of ML algorithms. Hence, the research communities need to consolidate their efforts in creating methods and tools that enable a holistic approach to safety of autonomous systems. This motivated the topic of our Dagstuhl Seminar - exploring the problem of engineering and safety assurance of autonomous systems from an interdisciplinary perspective. As a result, the discussions of achievements and challenges spanned over a broad range of technological, organizational, ethical and legal topics summarized in this document.