As organizations increasingly invest in AI upskilling to build a future-ready workforce, employees are expected not only to acquire AI knowledge but also to share it with coworkers, allowing individual learning to be transformed into collective organizational capability. However, employees who acquire AI knowledge do not always disseminate what they know, raising concerns that individually developed AI knowledge may fail to generate organizational value. Drawing on extended self theory, this study explores why self-directed learning of AI may unintentionally inhibit AI-knowledge sharing. We propose that self-directed learning of AI strengthens employees’ AI-knowledge ownership because personally acquired expertise is more likely to be perceived as “MINE” and incorporated into employees’ sense of extended self. This heightened ownership may subsequently increase employees’ AI-knowledge hiding. Furthermore, we propose that anticipated coworker negative responses to AI usage amplify this effect by increasing employees’ motivation to protect and retain their AI knowledge.