This study established cognitive vulnerability of anxiety symptoms among high school students. A total of 72 grade 9-11 students completed measures on levels of anxiety sensitivity (AS), selective attentional processing, and anxiety symptoms annually between 2016 and 2018. Latent class growth analysis (unconditional model) showed a four-class model: High (stable) (6.94%), low (stable) (11.11%), medium (decreasing) (61.11%), and medium (increasing) (20.83%). The conditioned model controlling for the physical-concerns dimension of AS and negative attentional bias demonstrated that a two-class model consisted of a low anxiety class (n = 59, 81.9%) and a high anxiety class (n = 13, 18.1%) provided the best fit for the data. Negative attentional bias is a significant factor related to the development of anxiety trajectories. Attentional bias modification to disengage from negative stimuli may serve as a potential target of intervention to reduce chronic anxiety among high school students.