I'm Sorry Driver, I'm Afraid I Can't Do That: Appraising the Safety of LLMs within Automotive Contexts

The paper evaluates the integration of large language models (LLMs) into automotive control tasks, focusing on safety assurance challenges. It identifies conceptual issues related to adapting general-purpose models to specific vehicle architectures and highlights engineering constraints from standards like ISO21448 and LLM-specific alignment problems from ISO/PAS8800. The findings underscore the need for improved safety mechanisms in deploying LLMs in real-time, safety-critical automotive environments, which is crucial for practitioners aiming to implement AI solutions in this sector.