User acceptance of personal AI agents: Benefits and risks of personal data use

Artificial intelligence (AI) technologies that rely extensively on personal data are increasingly integrated into everyday consumer contexts. While existing technology acceptance research has emphasized cognitive evaluations such as perceived usefulness and trust, data-intensive and opaque AI systems raise psychological challenges that are not fully captured by these frameworks. In particular, users often lack the information required to form stable trust judgments yet still make acceptance decisions under conditions of uncertainty. Drawing on Yamagishi's (1998, 1999) distinction between trust and assurance, and focusing on perceived assurance, this study reconceptualizes consumer acceptance of personalized AI as an affective–cognitive regulation process. We introduce perceived assurance, referring to users' perceptions that structural or institutional conditions make harmful behavior unlikely, alongside category trust, anthropomorphism, and perceived loss of autonomy. Using a scenario-based experimental design with video-based explanations, we examine how these cognitive, affective, and experiential factors jointly shape acceptance intentions toward a hypothetical personal AI agent. The findings highlight the importance of perceived assurance—acting as a structural risk-mitigation mechanism grounded in external safeguards—alongside experiential engagement in acceptance judgments, complementing traditional instrumental evaluations. They also reveal important boundary conditions for anthropomorphic design strategies and suggest that perceived reliance on AI does not uniformly undermine acceptance. By integrating perceived assurance into technology acceptance research, this study advances understanding of human behavior toward personalized, data-intensive AI systems and provides a foundation for future research on trust, perceived assurance, and autonomy in human–AI interaction.