YouTube to Automatically Label AI-Generated Videos

YouTube on May 27 announced it will automatically label AI-generated videos across the platform, shifting from a self-disclosure regime to platform-side detection. The announcement topped Hacker News for the day with 1,278 points and 769 comments, with the debate centring on detection reliability, false-positive rates for VFX-heavy human productions, and creator workflow impact.

The move is the most significant platform-level AI-content policy since Meta's labeling rollout, and lands as generative video tooling (Google Omni, Sora updates, Runway) has crossed a threshold where short-form AI clips are now routinely indistinguishable to viewers. YouTube has not disclosed the detection model architecture, raising the classic question of whether the label survives a quick re-encode or screen-capture-and-reupload.

Competitive context: the same week, NVIDIA AI publicised that the HaoAI Lab cut 5-second video generation from 25s on 8 Blackwell GPUs to 4.2s on a single Blackwell — open-sourced — making AI video production cheap and ubiquitous. Auto-labeling is YouTube's bet that platform-side disclosure is a more defensible policy than relying on creator honesty.

Watch: false-positive rates as creators report them, whether advertisers price labeled-AI inventory differently, and whether the EU AI Act treats YouTube's labels as a compliance artifact.