DF26 Benchmark: Humans and Detectors Both Near Random at Spotting AI Speech Videos

The new DF26 benchmark of 271 real and 2,420 synthetic public-speech videos shows both humans and top detectors perform near chance at identifying AI-generated videos, highlighting the need for robustness to distribution shift in modern generative models.

2026-09-10 ~ 2026-09-10 · 2 related posts

• DF26 benchmark: humans and detectors near chance on AI speaking videos — Severyn Shykula · 2026-09-10
• DF26 benchmark shows humans and top deepfake detectors are near random at spotting AI videos — ducha_aiki · 2026-09-10