AI Awareness and Personal Data Protection Practices in Bangladesh

Abstract Generative artificial intelligence creates opportunities for learning and professional work while requiring users to make decisions about sensitive information. This study examines whether self-reported AI awareness aligns with two distinct personal-data protection practices in Bangladesh: avoiding sensitive information in public AI tools and checking privacy or data-use information before sensitive tasks. A cross-sectional analysis of 385 adult respondents uses item-level distributions, paired endorsement contrasts, proportional-odds models and direct comparison of outcome-specific coefficients. The sample includes 242 law students alongside graduates, legal professionals and other respondents; it is not representative of all university students or the national population. Checking privacy information was endorsed by 284 respondents (73.8%), compared with 222 (57.7%) for avoiding sensitive input. Their within-person difference was 16.1 percentage points (95% bootstrap interval 9.6–22.6; Holm-adjusted p < .001). Self-reported understanding of privacy risks was endorsed by 275 respondents (71.4%), whereas understanding that AI services may collect, store or process user data was endorsed by 177 (46.0%). None of the eight focal awareness–practice coefficients received clear support after Holm correction; the joint comparison of associations across the two practices was inconclusive (χ²(4) = 1.83, p = .768). The findings distinguish self-reported understanding, reported protective practices and their uncertain adjusted relationships. Privacy calculus and protection motivation theory inform interpretation, but the questionnaire does not measure all components required to test either theory. Future research should document recruitment and evaluate privacy decisions through observed tasks or experiments involving specific AI services.