Generative artificial intelligence is rapidly entering various stages of undergraduate thesis writing, including topic selection, literature review, research design, data processing, text polishing, and thesis defense preparation. Unlike regular coursework, the undergraduate thesis carries attributes of degree conferral, academic training, and supervisor responsibility. Therefore, AI intervention is not merely a matter of efficiency tools but involves ethical issues such as authorship, originality, academic integrity, data privacy, boundaries of supervisor guidance, and university governance rules. Drawing on existing quantitative research frameworks on the relationship between ChatGPT academic use and academic integrity [1], this study focuses on the theme "Ethical Risks and Influence Mechanisms of Generative AI Intervention in Undergraduate Thesis Supervision" and constructs a structural model comprising perceived time-saving, thesis pressure, peer usage norms, supervisor clarity of AI guidance, academic integrity awareness, plagiarism risk perception, AI usage intensity, uncritical AI dependence, and responsible AI use. To demonstrate research design and paper writing style, this study sets up 386 undergraduate questionnaires as illustrative simulation samples and employs PLS-SEM for measurement model, structural path, and moderation effect testing. Results show that perceived time-saving, thesis pressure, and peer norms significantly increase AI usage intensity; AI usage intensity further increases the risk of uncritical dependence; academic integrity awareness, plagiarism risk perception, and supervisor clarity of AI guidance significantly reduce uncritical dependence; moreover, both academic integrity awareness and supervisor clarity of AI guidance weaken the positive effect of AI usage intensity on uncritical dependence. The findings suggest that universities should not respond to AI writing solely through "prohibition" or "detection," but should incorporate AI use disclosure, fact-checking, process-oriented supervision, phased documentation, and supervisor-student shared responsibility mechanisms into undergraduate thesis management.