Apple filed a motion to dismiss a class action lawsuit alleging the company illegally scraped millions of YouTube videos to train its AI models, arguing on July 2 that publicly available content does not qualify for legal protection against circumvention under the Digital Millennium Copyright Act. The lawsuit filed in April 2026 by the owners of h3h3Productions, MrShortGame Golf, and Golfholics, accuses Apple of deliberately circumventing YouTube’s anti-scraping protections and profiting substantially from the unlawful access. Apple’s response hinges on a narrow legal interpretation: because the videos are publicly viewable without password, payment, or authentication, YouTube’s technological measures do not “control access” to the works as Section 1201(a) of the DMCA requires.

The dispute centers on the Panda-70M dataset, which Apple researchers used to train STIV, a video generation model. According to Apple’s published research, Panda-70M functions as an index file mapping specific YouTube videos and clips by URL, video identifier, and timestamp, encompassing millions of videos harvested from the platform. The plaintiffs allege that Apple circumvented YouTube’s protections at scale to build this dataset, violating both the DMCA and the platform’s Terms of Service. Apple counters that the plaintiffs themselves made the content publicly available and that accessing public material does not constitute unauthorized circumvention.
Apple’s legal position is direct as per its motion as the company stated that “Plaintiffs allege that they posted audiovisual works to YouTube, and that any member of the public can see them there. No password. No payment. No lock. No key.” The argument proceeds that YouTube’s technological measures are defensive in nature, not access-controlling, because the videos are already freely viewable to anyone. By this logic, no circumvention has occurred, and the plaintiffs have failed to state a valid claim under DMCA Section 1201(a).
The three creators suing Apple command significant reach. h3h3Productions, created by Ethan Klein and Hila Klein, commands millions of followers across its main channel and the H3 Podcast, while MrShortGame Golf and Golfholics each boast hundreds of thousands of subscribers. These same creators have filed equivalent lawsuits against Meta, Nvidia, ByteDance, and Snap, signaling an organized legal front against multiple AI companies over unpaid use of creator content. The simultaneous actions suggest that large-scale scraping for AI training has become a focal point for the content creator economy.
The core legal question now rests on how courts interpret the interaction between public accessibility and technological protection measures under Section 1201(a). Plaintiffs argue that Apple’s deliberate circumvention of YouTube’s defenses, regardless of public visibility, violates the statute. Apple contends that the statute simply does not apply when content is already openly accessible. The distinction carries weight: if Apple’s reading prevails, any publicly posted content becomes effectively scrapable for AI training without copyright holder consent, provided no distinct access barrier exists. If the court sides with the creators, technological circumvention claims could survive even when material is public, potentially reshaping how AI companies source training data.
Apple’s filing reflects a broader tension between the tech industry and content creators over AI training datasets. Every major AI company is training on YouTube videos, making this dispute significant for the future of how AI companies must negotiate data rights directly with creators going forward. The plaintiffs described their position not as a narrow copyright grievance but as a challenge to the foundational funding model of generative AI: “not only unlawful, but an unconscionable attack on the community of content creators whose content is used to fuel the multi-trillion-dollar generative AI industry without any compensation.” Whether courts agree will determine whether large-scale unlicensed scraping remains viable or whether licensing becomes mandatory.
via MacRumors



