Deepfake detection is technically mature but legally unusable: no current system explains why audio is classified as synthetic. This project delivers forensic-grade explainability and reproducible benchmarks, unlocking adoption in forensics, law enforcement, and finance.