Opportunities and Risks of Generative AI in Model-Based Engineering of Railway Systems

The engineering of safety-critical systems demands rigorous assurance while facing growing requirements for automation and resilience. Traditional practices rooted in formal methods and model-based engineering (MBE) have provided the methodological backbone for decades, enabling compliance with international standards. At the same time, the increasing availability of operational data and computational resources has fostered a rapid expansion of artificial intelligence (AI) and machine learning (ML). Recent surveys indicate a paradigm shift from purely deterministic engineering towards hybrid approaches that combine data-driven adaptability with formal rigor. Against this background, the emergence of Generative AI and Large Language Models (LLMs) offers new opportunities to alleviate persistent limitations of MBE, including skill shortages, time-consuming processes, and difficulties in bridging natural language requirements with formal specifications. In this paper we review current research trends on the usage of generative AI and large language models in engineering safety-critical systems with a focus on railway verification and validation. We discuss how those techniques can complement and enhance MBE workflows. Potential applications include automated requirement formalization, model transformation, and test generation; however those applications also raise concerns about transparency and regulatory compliance. By setting such discussion at the intersection of modeling, verification, and emerging AI techniques, the paper highlights opportunities and risks, while honoring the legacy of Prof. Alessandro Fantechi, whose seminal contributions in railway research continue to inspire rigorous and trustworthy innovation.