Deepfake Awareness and Judgements about Digital Evidence in Bangladesh

Abstract Deepfake awareness, caution about authenticity and perceived need for professional training concern different aspects of digital evidence. This cross-sectional study uses 385 adult responses from Bangladesh to examine how these judgements coexist and whether reported awareness is associated with endorsement of stronger skills for legal professionals. We report item distributions, joint endorsement, paired contrasts and exploratory adjusted regressions. Caution about treating digital media as authentic without verification was endorsed by 298 respondents (77.4%), compared with 246 (63.9%) who endorsed stronger professional skills. The paired difference was 13.5 percentage points (95% CI 7.3 to 19.7; Holm-adjusted p < 0.001). Technical authentication when manipulation was suspected was endorsed by 242 (62.9%). Of the 216 who reported understanding that AI can generate realistic fabricated media, 143 also endorsed stronger professional skills. Adjusted associations of awareness, caution and perceived loss of trust in unauthenticated evidence with the skills rating were small and imprecise; no focal term met the multiplicity-adjusted threshold. The results distinguish a general expectation of verification from endorsement of specific professional responses. They support assessing provenance, authentication and evidentiary reasoning as separate tasks when designing training. Survey responses do not measure actual detection accuracy, trust in a specific record, or legal admissibility. The mixed sample, which includes 253 law students, does not establish national prevalence or the views of practising lawyers as a distinct population.