Prevalence-Conditioned Risk-Limiting Authorization of Deepfake Video Detectors under Dataset Shift: Aggregate Results and Reproducibility Code

Aggregate-only results and reproducibility code for a preregistered, prevalence-conditioned deepfake-detector authorization study across FakeAVCeleb, DF40, and VCF. The release includes negative untouched confirmation outcomes, controlled simulations, independent recomputation, and an explicitly post-result descriptive prevalence-sensitivity analysis. No source media, original paths, identities, row-level labels, crops, model weights, or per-item predictions are redistributed. Documentation and aggregate artifacts are CC BY-NC 4.0; original project-authored Python and PowerShell code is MIT.