Cybersecurity Behavior in AI Vibe Coding: Extending Technology Threat Avoidance Theory

Generative AI has rapidly reshaped software development, fueling “vibe coding,” in which developers increasingly rely on AI-generated code. While offering productivity benefits, this practice raises important cybersecurity concerns. Drawing on Technology Threat Avoidance Theory (TTAT), this study examines developers’ cybersecurity behavior in AI-assisted coding environments. Survey data from 358 participants with recent AI coding-assistant experience were analyzed using PLS-SEM. Security self-efficacy exhibited the largest positive effect on cybersecurity behavior, followed by perceived threat, while trust in AI had a smaller positive effect. AI transparency and perceived risk showed no significant direct effects. Perceived severity and susceptibility positively shaped perceived threat, supporting TTAT’s core threat-appraisal mechanism. The model explained 44.5% of the variance in cybersecurity behavior. PLSpredict provided evidence of out-of-sample predictive performance. The findings clarify TTAT’s applicability and theoretical boundaries in AI-mediated software development and offer practical implications for developers, organizations, and AI tool providers seeking safer AI-assisted coding practices.