Signal ChildSafeAds Shared Task 2026: Commercial Content in Child-Facing YouTube Videos

Summary

ChildSafeAds is a shared task focused on detecting commercial content in YouTube videos likely to reach children and teenagers, built from a dataset of 3,360 videos across 939 channels. Each instance pairs a SponsorBlock-flagged sponsor segment with its transcript, video and channel metadata, and any linked sales or service page. Participating systems are asked to determine what kind of offer is being promoted, assign product categories, and identify legal risk flags, with evidence divided into four cumulative access levels so results can be weighed against data-collection cost. The organizers found that 45.5% of videos in the dataset failed to properly use YouTube's in-platform 'Includes paid promotion' ad disclosure label. GPT-5.4 was used to produce labels after expert review and iteration on the taxonomy and prompts, while GPT-5.6-luna independently labeled the development set. The report describes the task, data, and evaluation methodology, with participating systems and results to be added in an updated version.