Disruptive technologies in the transport system
The development of disruptive technologies such as artificial intelligence (AI), autonomous and electric vehicles, and smart grids challenges legal frameworks to remain relevant, adaptive, and coherent over time. This is evident in the transport sector, where technological innovation has the potential to fundamentally reshape transportation, logistics, and urban design, or is already doing so. Against this backdrop, the present study examines the legal ramifications of such transformations by exploring whether AI-based tools can support forward-looking legal research. Specifically, it aims to identify and map potentially complex or unclear legal situations that may arise from the deployment of disruptive technologies. This study employs two large language models (LLMs), Microsoft 365 Copilot and Claude Sonnet 4.6, to generate possible future legal scenarios for the period 2026 – 2037, using a scenario-based research design primarily situated within a Swedish and EU legal context. Several methods for engaging with the LLMs are applied, including mixed prompting, Chain-of-Thought prompting, the RACE framework, and meta-prompting. The results show that both LLMs kept well to the prompt instructions regarding output format, including the integration of future scenarios, external drivers, and potentially challenging legal themes. Additionally, while differences between the responses of the two LLMs were observed, these were primarily differences of degree rather than of kind. The overall conclusions suggest that both LLMs demonstrate potential usefulness for further research and may serve as exploratory tools in legal research as a point of departure. However, although the scenarios produced by the LLMs were grounded in contemporary legal research “hot topics”, a more thorough examination of the validity of their claims is required to support well-founded decisions regarding future research. Finally, it is recommended that researchers should inter alia continue refining their “prompt literacy” while still relying on human judgment, especially for critically evaluating the validity and robustness of LLM-assisted research.