The current safety evaluation standards lack guidance on the alignment design specific to lower-grade highways, such as four-lane and non-standard highways. To address this issue, a method for assessing the driving safety of mountainous highways with low index and complex alignment was proposed. The method involved building a human-vehicle-road system simulation using the CarSIM/TruckSIM simulation software platform, selecting representative car models, and using realing vehicle driving data as simulation parameters to carry out virtual vehicle driving tests. Four categories of indicators, including speed, lateral stability, driver handling, and speed regulation, were used to examine and analyze the compliance of the tested highway’s geometric design, speed coordination, driving operation load, identification of accident-prone sections, and safety of continuous uphill sections. The feasibility of the “human-vehicle-road” simulation method was verified by running virtual simulation driving tests similar to the real-vehicle test, and one four-lane highway and one non-standard highway were selected as examples. The results demonstrate that this method can effectively identify alignment design combinations that are below the limit and inconsistent, as well as road sections with high driver handling loads and poor balance. It can also identify dangerous road sections, corresponding vehicle accident patterns, critical safety speeds, and positions and widths where heavy-duty trucks deviate from the travel lane when passing through hairpin curves. Based on the driving/accident conditions of dangerous road sections, targeted geometric parameter improvement measures can be proposed to enhance the trafficability and stability of vehicles on ordinary mountainous highways with low index and complex alignment, ultimately improving the quality of alignment design and driving safety for low-index highways in mountainous areas.