Judge lets YouTubers' anti-circumvention claim against Apple's AI video scraping proceed

On October 9, 2026, Judge Richard Seeborg of the US District Court for the Northern District of California denied Apple’s motion to dismiss in Ted Entertainment, Inc. v. Apple, Inc. (case 3:26-cv-02936). The plaintiffs, Ted Entertainment and creator Matt Fisher, sue for a putative class of YouTube creators under the anti-circumvention provision of the Digital Millennium Copyright Act, 17 U.S.C. 1201(a), rather than for copyright infringement.

Their amended complaint alleges that Apple scraped millions of YouTube videos to “feed, train, improve, and commercialize” its generative text-to-video AI models. According to the order’s summary, Apple started from the public HD-VILA-100M dataset to build its own Panda-70M dataset, then used tools such as yt-dlp to defeat YouTube’s rolling cipher, rotated IP addresses to avoid blocking, and regenerated links and reused proof-of-origin tokens to get around session-bound URLs and CAPTCHAs.

Apple made three arguments: that YouTube’s measures are use controls under section 1201(b), not access controls under 1201(a); that they do not operate in the ordinary course; and that its conduct was not “circumvention”. The court rejected all three at the pleading stage. Citing Ninth Circuit precedent in MDY v. Blizzard and Disney v. VidAngel, it held that a measure can be both a use and an access control, need not be circumvention-proof, and qualifies if it gives “some degree of control over access” to a work. The alleged rolling cipher, IP blocking and session-bound URLs were enough.

Why it matters: fair use rulings since 2025 have made infringement claims over AI training harder to win, and creators are turning to how training data was obtained. Section 1201 has no fair use defense, so a claim that a developer broke through a platform’s technical protections can survive even where training itself might be lawful.

What it does not show: the ruling accepts the complaint’s allegations as true and decides nothing about whether Apple actually circumvented anything. The order does not discuss AI training; that context comes from the complaint. The class is not certified.