Lawsuit: Apple allegedly used YouTube videos for AI Apfelpatient

A class-action lawsuit accuses Apple of using millions of YouTube videos to train an AI model. Amazon and OpenAI are also implicated.

Apple's AI ambitions are once again facing legal turmoil. Three YouTube creators have filed a class-action lawsuit against Apple, alleging that the company circumvented YouTube's anti-scraping mechanisms to use video footage for training its own AI model. This case joins a growing body of litigation surrounding AI and copyright and could have far-reaching consequences for the entire industry.

At the heart of the lawsuit is a dataset called Panda-70M, which contains references to millions of YouTube videos. According to a 2024 study titled "STIV: Scalable Text and Image Conditioned Video Generation," Apple researchers used this dataset to train an AI-powered video generation model. The plaintiffs—Ted Entertainment, Matt Fisher, and Golfholics—claim that their content appears over 500 times in the dataset and are seeking representation for all affected YouTube creators (via MacRumors).

How the data set works

The Panda-70M dataset does not contain the videos themselves, but functions as a kind of index: It references specific YouTube videos and individual clips within them via URL, video ID, and timestamp. A single video can be divided into numerous clips, each of which is treated as a separate training sample.

The plaintiffs argue that extracting each individual clip requires separate access to the source video on YouTube – and that each of these accesses constitutes a separate circumvention of YouTube's copy protection mechanisms. The lawsuit is thus based on the Digital Millennium Copyright Act (DMCA), which criminalizes the circumvention of technical protection measures.

Not only Apple is affected

The plaintiffs have filed identical class-action lawsuits against Amazon and OpenAI. Both companies are accused of using the same Panda 70M dataset for their own AI training processes. The case is thus part of a broader legal debate about whether and under what conditions publicly available content on the internet may be used for AI training.

What the plaintiffs are demanding

The lawsuit aims at several points: The plaintiffs are seeking certification as a class action, a court declaration that Apple knowingly circumvented protective measures, statutory damages at the maximum amount per violation, a permanent injunction against the further use of the affected content, and reimbursement of legal fees. They are also requesting a jury trial.

Apple's difficult position

This case is particularly sensitive for Apple because the company has consistently positioned itself as a privacy-conscious player in its communications surrounding Apple Intelligence. The allegations of using third-party content for AI training without permission stand in stark contrast to this image.

Apple has not yet commented publicly on the lawsuit. The outcome of the proceedings depends largely on whether the court classifies the use of the Panda 70M dataset as copyright infringement – or whether Apple can argue that the public accessibility of the videos justifies their use for research purposes. The result is likely to be groundbreaking not only for Apple, but for the entire AI industry. (Image: Shutterstock / NorthSky Films)