Apple Hit With Class-Action Over AI Training Scrapes

Three YouTubers say Apple scraped their videos to train its AI—without permission.

A trio of creators behind h3h3 Productions, MrShortGameGolf, and Golfholics has filed a class-action lawsuit alleging that Apple violated the Digital Millennium Copyright Act by scraping copyrighted YouTube videos to train Apple’s generative AI models. The suit contends Apple circumvented YouTube’s “controlled streaming architecture,” giving the tech giant access to vast video content beyond what everyday users can view, and that this access fed Apple’s AI products. The creators argue that Apple’s financial success rests, in part, on content they produced and that this training approach deprives them of control and compensation.

The complaint, highlighted by MacRumors in reporting that first spotted the filing, adds that the YouTubers’ material remains publicly available on YouTube, but that Apple’s automated data-gathering went beyond ordinary user activity. It’s part of a broader wave: creators have separately targeted other major tech players over alleged AI-training practices. OpenAI and Microsoft, for instance, faced claims that copyrighted articles from the New York Times informed their chatbots, while Perplexity has been pressed by Reddit and Britannica over copyright and trademark concerns. In addition, Apple has already been named in another DMCA-related action brought by neuroscience professors who allege their copyrighted work was used without permission.

Industry observers say the suit underscores a rapidly evolving, high-stakes debate over what counts as permissible data use in AI training. The DMCA framework, and specifically the anti-circumvention provisions, is central to the disagreement: if a company bypasses platform-level protections to collect data at scale, does that cross a line that ordinary users cannot cross? The plaintiffs’ argument hinges on the claim that Apple’s approach exploited a “controlled streaming architecture” to access content far beyond what licensing agreements or fair-use considerations would anticipate.

Practitioner insights are emerging from creators, platforms, and AI developers. First, data sourcing remains the most consequential choke point: without clear licensing pathways or robust fair-use precedents, AI developers risk costly litigation and potential licensing bottlenecks that could slow product iterations. Second, platform policies and terms of service will matter more than ever—whether a company can demonstrate that it operated within or beyond those rules could shape outcomes, even if the broader legal questions remain unsettled. Third, for content creators, the case reinforces the risk that monetizable value sits in the hands of those who control access to large-scale data, not just the end-user revenue from views. Finally, this development could spur industry-wide shifts toward licensed datasets, creator-consent frameworks, or on-device training constraints to minimize reliance on scraped content.

The legal process remains uncertain in detail—the report does not reveal the court, filing date, or docket, complicating predictions about how quickly the case might proceed or what discovery will uncover. Still, the complaint signals that the clash over AI training data is moving from rumor and advocacy into formal litigation with real financial and competitive stakes. If Apple or other tech giants face adverse rulings, it could sharpen incentives for more transparent data licensing, stronger creator protections, and new norms for how AI models are trained on publicly available content.

Sources

Three YouTubers accuse Apple of illegal scraping to train its AI models