<![CDATA[The rise of generative AI (Gen-AI) tools has blurred the ethical boundaries of academic writing, shifting the definition of plagiarism from traditional methods to AI-based plagiarism (AI-giarism). English as a Foreign Language (EFL) students are the most vulnerable people to fall into this practice due to second language writing anxiety, which drives dependency on AI as a shortcut. This study aims to examine the levels of knowledge, attitudes, and self-reported AI-based plagiarism practices among EFL students at a pesantren-based higher education institution. Employing a quantitative cross-sectional survey design, data were collected using a validated 51-item questionnaire adapted from Phyo et al. (2023), distributed to 99 English education students at a pesantren-based higher education in East Java, Indonesia, via Google Forms. Data were analyzed using descriptive statistics, Spearman’s correlation, multiple linear regression, and binary logistic regression. Result revealed a moderate level of knowledge (mean = 9.63/17; 56.6%), with the greatest weakness in understanding that AI-generated text passing Turnitin still constitutes plagiarism (only 33.3% correct). Attitudes rejecting plagiarism were at a moderate level (mean = 91.24/135; 67.6%), with the self-justification subscale scoring the lowest (55.5%). A total of 50.5% of respondents had engaged in at least one act of AI-giarism. Knowledge score emerged as the sole significant independent predictor of AI plagiarism behavior (AOR = 0.764;95% CI [0.62-0.94]; p = 0.012). These findings reveal the “Pesantren Paradox”: a conflict between strong religious moral capital and limited digital-ethical literacy, which creates a loophole for unintended AI-giarism.]]>
