Slow-moving active landslides pose a significant threat to infrastructure, particularly along highways that cross tectonically active mountain belts. In high-altitude mountainous regions, identifying and assessing such landslides is difficult due to limited accessibility, extensive spatial coverage, a lack of versatile detection methods, and the inherently complex behavior of mass movements. In this research, we utilized Sentinel-1 Synthetic Aperture Radar data spanning from 2015 to 2024 to identify active landslides along two key segments: the alternate route of the Karakoram Highway (Chitral-Gilgit) and the main Karakoram Highway section (Gilgit-Khunjerab). By integrating phase gradient stacking with a deep learning network, we detected a total of 1,037 active landslides across the study area. Subsequently, timeseries InSAR analysis was conducted on selected large-scale landslides to determine their displacement rates and deformation patterns, which revealed line-of-sight velocities as high as -81 mm per year. The reliability of these detections was confirmed through cross-validation with Google Earth imagery and previously published landslide inventories for the Karakoram Highway region. The findings illustrate the spatial distribution of active landslides within the uplifting mountainous terrain, identify zones of potential instability, and provide valuable insights for hazard mitigation and risk assessment, particularly, for less-monitored economic roadways located in orogenic settings.