Prevalence-induced concept change (PICC) occurs when reduced category prevalence increases the likelihood that ambiguous stimuli are classified as belonging to the now-minority category. PICC has been observed across perceptual and social domains and persists despite instructions or incentives to suppress it. However, other findings suggest its expression is instead shaped by social context. If AI models are to be treated as theories of cognition, they should exhibit PICC as well. We show that a standard AI model for sequential learning, trained on dynamic category distributions, does not display PICC, instead favoring the now-majority category. This opposite-PICC effect suggests that simple sequential learning may be insufficient to produce PICC. Additional mechanisms, such as structured priors, contextual sensitivity, or internal feedback, seem necessary for its emergence. Our findings contribute to the understanding of PICC and its implications for categorization theories, AI-driven decision-making, and the role of media exposure in shaping social perceptions.