James v. Together Computer, Inc.

On 6 November 2025, James filed a class action complaint against Together AI. Together AI is a San Francisco-based company that sells cloud platform access to help other entities train, fine-tune, and deploy generative AI models. In April 2023, Together AI assembled and published the RedPajama dataset, which contained a deduplicated copy of the Books3 dataset, a pirate library allegedly containing approximately 196,640 books. The Complaint states that the RedPajama dataset contained a subset called "Books" or "RedPajama-Books" that was actually a copy of the Books3 dataset. Together AI hosted the dataset on its own GitHub and Hugging Face pages and actively promoted its use, including through a $1 million prize competition encouraging third parties to fine-tune models using RedPajama data. Between April and October 2023, the dataset was downloaded over 190,000 times. Books3 was eventually removed from RedPajama due to reported copyright infringement.

The Complaint brings two causes of action. The first is direct copyright infringement under 17 U.S.C. § 501, alleging Together AI violated the plaintiff's exclusive rights by downloading, copying, storing, and republishing copyrighted works within RedPajama. The second is contributory copyright infringement, alleging Together AI had actual or red-flag knowledge that the dataset contained pirated works, and materially contributed to third-party infringement by making the dataset freely available and actively encouraging its download.

This case is notable because Together AI is not accused of training its own models on the infringing data, but rather of assembling and distributing the infringing dataset itself. The contributory infringement claim targets Together AI's role as a dataset publisher and promoter, arguing that making copyrighted training data freely available and incentivising its download constitutes material contribution to downstream infringement by third parties. This positions Together AI as an infrastructure provider in the AI training pipeline rather than an end user of the copyrighted works.

The Complaint seeks statutory damages, actual damages, restitution of profits, a declaration of wilful infringement, and destruction of all infringing copies. No specific dollar amount is stated.

Case referred to private ADR on 19 February 2026.