Two copyright actions filed against OpenAI in the Southern District of New York in August, a putative class action led by textbook author Michael Sullivan and a separate action brought by wikiHow, have been consolidated into In re OpenAI, Inc. Copyright Infringement Litigation, multidistrict litigation (MDL) No. 3143, before Judge Sidney Stein.
They join proceedings involving authors, news organisations and digital content creators, including the Authors Guild, George R.R. Martin, John Grisham, Paul Tremblay, Sarah Silverman, The New York Times and YouTube creator David Millette.
On 1 September, the United States filed a Statement of Interest, applicable to all matters in the MDL, addressing a central question in the litigation: whether copying copyrighted works to train large language models constitutes fair use.
Sullivan class action
On 14 August 2026, seven plaintiffs led by mathematics author Michael Sullivan filed a putative class action on behalf of textbook authors against OpenAI and Microsoft. The plaintiffs allege that OpenAI obtained their works from pirated shadow libraries such as LibGen, ran a script to strip copyright management information from them, and used them to train its GPT models.
The complaint places particular emphasis on potential market harm. The plaintiffs argue that the textbook market differs from the market for trade books because textbooks are typically selected by instructors or institutions rather than by the students who use them. They contend that AI-generated educational content could be substituted for textbooks, diluting demand for the plaintiffs’ works and undermining the market for producing updated educational materials. The allegations against Microsoft include its provision of Azure infrastructure used for training. The plaintiffs plead direct, vicarious and contributory infringement, alongside a claim under section 1202(b) of the Digital Millennium Copyright Act (DMCA).
wikiHow complaint
On 21 August 2026, wikiHow, Inc. filed its own complaint against nine OpenAI entities. The publisher asserts 1,211 copyright registrations covering 11,211 how-to articles and alleges infringement at three stages: ingestion into training data, retrieval when responding to users through retrieval-augmented generation, and reproduction in ChatGPT outputs.
wikiHow also alleges that OpenAI’s crawlers continued to access its site after it added robots.txt directives instructing them not to do so, and that its approaches to OpenAI did not result in a licensing agreement. Its DMCA claim concerns extraction tools that separate article text from surrounding bylines and copyright notices.
US Statement of Interest
On 1 September 2026, the Department of Justice filed a Statement of Interest under 28 U.S.C. § 517, expressed to apply to all matters in the MDL. The United States argues that copying written works to train a large language model is ‘extraordinarily transformative’ and constitutes fair use. It also warns that requiring licences could concentrate model development among the largest technology companies.
The filing addresses the differing reasoning in the two Northern District of California rulings of June 2025. It endorses Judge Alsup’s analysis of training in Bartz v Anthropic and rejects as ‘deeply flawed’ the market dilution theory discussed by Judge Chhabria in Kadrey v Meta. Both courts found the training uses before them to be fair use, but differed on the potential significance of competition from AI-generated works.
The government also challenges the Copyright Office’s analysis of market harm in its May 2025 report.
As Sasha S. Rao and Ivy Clarice Estoesta of Sterne Kessler observe, it ‘sidesteps the question of whether training on pirated materials is fair use’. The distinction between acquiring works and using them for training therefore remains significant to the allegations in Sullivan.
The government’s position is also consistent with the broader US policy of supporting AI development and reducing barriers to innovation. This position is particularly notable against the renewed debate over the pace and oversight of frontier AI development. In recent weeks, leaders of several frontier AI companies have themselves called for stronger safety oversight, independent evaluation and, in some cases, a slowdown in the development of increasingly capable systems.
The tension between encouraging rapid AI development and responding to growing calls for greater oversight is likely to remain an important feature of US AI policy. It will also be interesting to see whether the United States files statements of interest in other AI copyright cases as these issues continue to develop before the courts.
Further case information is available through the Deep Lex Disputes Tracker.
Sources
- Sullivan v OpenAI Foundation, No. 1:26-cv-06966 (S.D.N.Y.), Class Action Complaint, 14 August 2026.
- wikiHow, Inc. v OpenAI, Inc., No. 1:26-cv-07171 (S.D.N.Y.), Complaint, 21 August 2026.
- In re OpenAI, Inc. Copyright Infringement Litigation, MDL No. 3143 (S.D.N.Y.), Statement of Interest of the United States, 1 September 2026.
- U.S. Copyright Office, Copyright and Artificial Intelligence, Part 3: Generative AI Training, pre-publication version, May 2025.
- Sasha S. Rao and Ivy Clarice Estoesta, ‘IP Hot Topic: The DOJ Chimes in on Fair Use and AI’, Sterne Kessler.
- Paulin Kola and Lily Jamali, ‘Anthropic boss Dario Amodei calls for AI development to slow down’, BBC News, 12 September 2026.
- Deep Lex, AI Disputes Tracker.
This article is for educational and research purposes only and does not constitute legal advice.