Introduction: As AI evolves into agent-based forms, AI threat perception has become a defining workplace concern. Informed by Cognitive Appraisal Theory and Protection Motivation Theory, this cross-sectional study investigates how AI threat perception relates to post-adoptive AI agent engagement via dual paths of AI self-efficacy and defensive motivation. Methods: = 13) first grounded scale adaptation. Structural equation modeling was then conducted on data collected from 387 knowledge workers, followed by multi-group analysis. Results: The structural model demonstrates that AI threat perception is indirectly associated with post-adoptive engagement through both mediators, exhibiting an indirect-only pattern. Multi-group analysis further reveals demographic and contextual boundary conditions, showing that higher education significantly amplifies defensive motivation (the link is stronger among postgraduates), while exploratory career-stage and regional comparisons indicate that the self-efficacy pathway is stronger among early-to-mid career professionals and workers in eastern regions. Discussion: The study highlights the concurrent coexistence of capability beliefs and defensive intentions in post-adoptive use, while pointing to the need for longitudinal research to trace their temporal dynamics. These findings refine our understanding of technology threat perception and inform human-agent collaboration.