IntroductionAs artificial intelligence (AI) becomes increasingly embedded in organizational workflows, understanding how human–AI interactions shape employees’ knowledge territorial behavior has become a critical issue in digital transformation. Drawing on conservation of resources theory, this study examines the differential effects of enhanced and impeded human–AI interaction on employees’ knowledge territorial behavior, with organizational dehumanization as the underlying psychological mechanism and paradoxical leadership as a boundary condition.MethodsWe tested the hypotheses across two studies. Study 1 used a scenario-based experiment to establish the causal effects of different forms of human–AI interaction. Study 2 employed a two-wave field survey to validate the full moderated mediation model in a real-world organizational context.ResultsEnhanced human–AI interaction was negatively associated with knowledge territorial behavior, whereas impeded human–AI interaction was positively associated with such behavior. Organizational dehumanization significantly mediated these relationships. In addition, paradoxical leadership moderated these effects: higher levels of paradoxical leadership strengthened the negative effect of enhanced human–AI interaction on organizational dehumanization and attenuated the positive effect of impeded human–AI interaction on organizational dehumanization.DiscussionThese findings reveal the double-edged effects of human–AI interaction on employees’ knowledge territorial behavior. This research extends the literature on the antecedents of knowledge territoriality, highlights organizational dehumanization as a key psychological mechanism, and identifies paradoxical leadership as an effective managerial approach for facilitating human–AI collaboration in modern organizations.