Bridging the economic safety gap: Leveraging Generative AI for Safe Automation

The rapid advancements in AI and autonomy have accelerated innovation in industrial automation, yet safety standards and certification processes often struggle to keep pace. This widening gap, the Economic Safety Gap, highlights how organizations risk missing the economic benefits of emerging AI-driven tools when strict safety requirements cannot be efficiently fulfilled or validated. With the increasing prevalence of adaptive systems, collaborative robots, and data-intensive analytics, traditional safety methods remain time-consuming and costly, creating a pressing need for more agile, AI-assisted safety engineering. In response, recent research demonstrates how generative AI, particularly large language models, can aid in hazard identification, risk assessment, and the generation of formal safety arguments. By automating repetitive tasks and systematically identifying potential weaknesses, generative-AI-based solutions can accelerate safety analyses across various industrial domains. However, these approaches also face reliability concerns, including hallucinations and domain incompleteness, underscoring the enduring importance of expert oversight. This paper presents a proof-of-concept assistant that integrates generative AI into the Hazard Analysis and Risk Assessment, demonstrating initial efficiency gains while maintaining required safety standards.