September 5, 2025, a few days after Anthropic reportedly paid $1.5 billion to settle claims with copyright holders and their counsel over the use of pirated books to train AI systems, two authors who identify themselves as copyright holders filed a proposed class action in a U.S. court accusing Apple of training its artificial intelligence systems on large volumes of book text without obtaining permission or providing compensation. The complaint centers on the "unauthorized sources of training data" and the resulting copyright infringement risks.
The complaint alleges that Apple collected and used e‑book and book content for its AI products — including the frequently promoted "Apple Intelligence" — and that such content did not come from authorized publishers or library licenses but from channels the plaintiffs characterize as "pirated" or otherwise unlicensed. The plaintiffs describe specific instances in which the model produced passages or sentences highly similar to original works, suggesting the training data included identifiable book text.
The suit asks the court to certify the case as a class action to represent all similarly affected authors, copyright holders and publishers and to pursue damages. The remedies sought include disclosure of Apple's training data sources and methods, deletion or cessation of use of allegedly infringing data, and compensation for economic losses stemming from Apple's commercial use of the authors' works.
The case echoes earlier litigation brought against other technology companies. In their filings, the plaintiffs' lawyers list the types of evidence they say will support their claims: fingerprints in training samples, text overlaps between AI outputs and original works, and Apple's public emphasis on the scale and diversity of its training data. The plaintiffs stress that allowing tech companies to use copyrighted books for model training without permission would cause long‑term harm to the traditional publishing industry and independent authors.
The plaintiffs seek procedural orders such as early preservation of evidence and broad discovery aimed at compelling Apple to disclose documents related to training data, data suppliers and internal model evaluations. The complaint also cites provisions of U.S. copyright law, asserting direct and secondary liability and asking the court to decide whether the use qualifies as fair use.
As of 9to5Mac's report, Apple had not publicly responded to or commented on the proposed class action. If the court accepts the case and grants class certification, it could become a significant precedent for how the United States evaluates the legality of large‑scale training data and could further affect Apple's commercial strategies for AI product development and content licensing.