
On April 3, 2026, Ted Entertainment Inc., content creator Matt Fisher and Golfholics Inc. filed a class action lawsuit against Apple Inc. in the U.S. District Court for the Northern District of California.
The lawsuit alleges Apple secretly downloaded millions of copyrighted YouTube videos to train a generative AI video model, bypassing the platform's security protections along the way. The complaint brings a single claim under the Digital Millennium Copyright Act, the federal law that prohibits bypassing digital security measures protecting copyrighted material.
The creators behind the lawsuit
Ted Entertainment Inc manages h3h3 Productions and H3 Podcast Highlights, two YouTube channels with more than 5,800 original videos and roughly 2.6 million subscribers, the complaint claims. The lawsuit alleges 438 of those videos appear in a dataset central to the case. Matt Fisher runs the golf instruction channel MrShortGame Golf with more than 500,000 subscribers, and Golfholics Inc operates a golf-focused channel with more than 130,000 subscribers. The complaint claims eight and 62 of their respective videos appear in the same dataset.
The three plaintiffs say they did not authorize Apple to download or use their content. They filed the lawsuit on behalf of a proposed nationwide class of YouTube creators whose videos allegedly ended up in Apple's AI training pipeline without their knowledge.
How Apple's own research paper allegedly exposed the scraping
In December 2024, 14 Apple employees coauthored a peer-reviewed academic paper titled "STIV: Scalable Text and Image Conditioned Video Generation." The paper identifies a dataset called Panda-70M as a training data source for Apple's generative AI video model, the complaint claims.
Panda-70M, originally compiled by Snap Inc., allegedly draws from roughly 3.1 million YouTube videos broken into about 70.8 million clips. The dataset does not contain actual video files but provides links to YouTube content. To use it for AI training, a company would need to visit YouTube and download every referenced video directly, the lawsuit claims.
Five layers of YouTube security at the heart of the case
The complaint describes five technological protection measures YouTube uses to lock down its video files:
- Rolling cipher encryption that scrambles video file URLs
- IP-based blocking and rate limiting that restricts high-volume automated requests
- Session-bound streaming URLs that expire quickly
- CAPTCHA challenges triggered by automated activity
- Proof-of-origin cryptographic tokens that verify authorized access
The lawsuit alleges Apple deployed automated video-downloading tools, specifically naming the open-source program yt-dlp, to crack the encryption. To dodge detection, Apple reportedly rotated IP addresses through virtual machines and renewed expired session credentials programmatically. The complaint also claims Apple distributed requests across multiple machines to avoid CAPTCHA triggers and replicated YouTube's origin verification tokens.
YouTube's terms of service expressly prohibit scraping, unauthorized downloading and bulk extraction of content. YouTube CEO Neal Mohan publicly called unauthorized scraping a violation of both creators' rights and the platform's terms of service.
What the lawsuit seeks
The plaintiffs ask the court to certify the case as a class action covering all U.S.-based YouTube creators whose videos Apple allegedly scraped through the Panda-70M dataset.
The complaint seeks the maximum statutory damages allowed under the DMCA, an injunction blocking Apple from continuing to use the allegedly scraped material and a declaration that Apple willfully circumvented YouTube's protections.
What this means for YouTube creators
The same plaintiffs filed similar lawsuits against Meta, Nvidia, ByteDance and Snap over related AI training practices. This case remains in its early stages with no settlement, no claims process and no money available. The lawsuit is pending in federal court in San Francisco. Apple has not publicly responded to the allegations.
.png)







.webp)
.webp)
.webp)

.webp)
.webp)
.webp)
.webp)



