Research on the construction of an integrated safety management system for intelligent connected vehicles

Large language models (LLMs) have fully entered the deep vehicle integration phase, achieving in-depth synergy with the automotive industry across core scenarios such as intelligent cockpit interaction, autonomous driving decision-making, and vehicle-wide intelligent control, thus driving the industry toward a wave of intelligent transformation. However, traditional automotive safety management frameworks are mostly built around mechanical hardware and basic electronic systems, exhibiting significant disconnects from the algorithmic black-box characteristics and data-driven logic of automotive AI technologies. A single safety system struggles to cover the compound risks arising from cross-domain collaboration across the "perception-decision-execution" chain, while fragmented management not only pushes up enterprise compliance costs but also creates barriers to data flow, seriously restricting safety management efficiency. To address this, this paper establishes a research framework for an integrated management system for intelligent connected vehicles (ICVs) encompassing functional safety, intended functional safety, and AI safety. Grounded in existing functional safety standards, the framework fully considers the full-lifecycle characteristics of ICVs from R&D and testing to operation and maintenance, providing systematic support for the safe development of the industry.