A developer using the handle hashfunction published metadata for approximately 5.6 billion public TikTok videos on Hugging Face, spanning July 2014 through October 2026, totaling 460 GB of Parquet files available for free under a CC BY-NC 4.0 license. The dataset contains captions, hashtags, sound IDs, on-screen text, TikTok Shop product and seller IDs, and engagement metrics such as views, likes, comments, shares, saves, and downloads, but no actual video footage. Access to TikTok’s private mobile API was achieved without login or account by generating spoofed device identities and reverse-engineered request signatures. DataSocial, the account behind the release, routes commercial use, creator profiles, and daily updates to datasocial.ai, where the scraper source code is sold for 1,699 dollars. TikTok’s terms explicitly prohibit automated scraping, and this release comes amid ongoing legal battles over data collection for AI training purposes.
Source: Read the original article
