Chicken Soup for the Soul, LLC v. Anthropic PBC
Complaint
Issue: Whether the unauthorized downloading and reproduction of copyrighted books from shadow-library repositories (including LibGen, Z-Library, Books3/The Pile, and Anna's Archive) to train and optimize commercial large language models constitutes willful copyright infringement under the Copyright Act, actionable by the copyright owner against multiple AI developers including Anthropic, Google, OpenAI, Meta, xAI, Perplexity, Apple, and NVIDIA.
Plaintiff Chicken Soup for the Soul, LLC — owner of registered copyrights in hundreds of titles in its namesebook franchise — filed this complaint against thirteen defendants, alleging that each obtained pirated copies of its works from shadow libraries and made additional unlicensed reproductions during data ingestion, preprocessing, and iterative model training. The complaint traces the alleged infringement to a single origin point: an OpenAI employee's 2018 download of LibGen, which it contends set an industry template that all other defendants subsequently followed, either through The Pile/Books3 or through independent access to additional shadow libraries. Plaintiff expressly declined to proceed as a class action, invoking its Seventh Amendment right under *Feltner v. Columbia Pictures* to jury-determined statutory damages — noting that a pending class settlement in a related Anthropic action would yield roughly $3,000 per work, approximately 2% of the Copyright Act's $150,000 statutory ceiling for willful infringement. Plaintiff seeks damages, permanent injunctive relief, and all other available remedies.
This complaint is notable for framing industry-wide AI training practices as a coordinated, cascading pattern of willful infringement rather than isolated conduct, and for the plaintiff's deliberate rejection of class-action treatment as a mechanism it characterizes as systematically undervaluing individual copyright claims against AI developers. If litigated to verdict, it could produce the first jury-assessed statutory damages award — potentially at the willful-infringement ceiling — against multiple major AI companies for training-data copyright claims, establishing a damages benchmark that would significantly complicate the class settlement framework currently emerging in related litigation.