The consolidated copyright litigation against OpenAI (In re OpenAI, Inc. Copyright Infringement Litigation, MDL No. 3143) has entered a critical phase following a series of rulings that largely favour the Plaintiffs who are authors, publishers, and news organisations including The New York Times.
Motion to dismiss denied
On 27 October 2025, Judge Sidney Stein rejected OpenAI's attempt to dismiss output-based infringement claims. The court found that ChatGPT-generated summaries of George R.R. Martin's novels were "substantially similar" to the originals because they "parrot[] the plot, characters, and themes." Fair use remains unresolved.
"Download claim" survives; Microsoft remains in the action
Judge Stein also refused to strike the Plaintiffs' theory that OpenAI's downloading of books from shadow libraries like Library Genesis constitutes infringement separate from training. This issue also formed part of the claims Bartz v. Anthropic, a case that settled earlier this year for US$1.5 billion. The court also denied Microsoft’s motion that the class action complaint only include models trained by OpenAI, and not those trained by Microsoft nor those which were used in Microsoft products eg Office and Co-Pilot. Microsoft therefore asked the court to exclude GPT-4o and GPT-4o mini from the scope of the class action complaint. That request was denied on 27 October 2025 thus Microsoft may now face discovery over a large volume of its products and services.
20 million ChatGPT logs to be produced?
Magistrate Judge Ona Wang ordered OpenAI to hand over 20 million de-identified user conversation logs. OpenAI has sought reconsideration, warning that "anyone in the world who has used ChatGPT" now faces potential disclosure of their conversations. The company argues over 99.99% of logs are irrelevant (ECF 742).
Privilege waived over LibGen deletion?
In a ruling on 24 November, Judge Wang found OpenAI waived attorney-client privilege over communications about deleting the Books1 and Books2 datasets (sourced from Library Genesis). Judge Wang found OpenAI's shifting positions on why it deleted the pirated book datasets, first citing "non-use," then claiming all reasons were privileged, constituted waiver.
OpenAI must produce these documents by 8 December and make its in-house lawyers available for deposition by 19 December.
Key Developments (incl. links to documents)
Date
Event
Doc (ECF)
27 Oct
Motion to dismiss denied; motion to strike "download claim" denied
7 Nov
20 million ChatGPT logs ordered to be produced
18 Nov
The Plaintiffs seek inspection of base/intermediate models
24 Nov
Privilege waived over LibGen deletion communications
24 Nov
30(b)(6) deposition topics dispute resolved in plaintiffs' favour
What's Next?
The discovery deadline is 26 February 2026. The Plaintiffs are also seeking to test OpenAI's "base" models, versions before consumer-facing filters were applied, arguing they may more readily regurgitate memorised training content. OpenAI has refused, but the Plaintiffs noted that the company routinely provides base model access to outside researchers.
Follow case updates and access the docket here: https://www.deep-lex.com/disputes/in-re-openai-inc-copyright-infringement-litigation
This post is for informational purposes only and does not constitute legal advice.