An updated complaint revises claims against Microsoft and drops one against OpenAI, as a landmark copyright case reshapes how AI companies use media content
The New York Times has filed an amended complaint in its copyright infringement lawsuit against OpenAI and Microsoft, adjusting the legal strategy it first deployed when it filed the original case on December 27, 2023, in the US District Court for the Southern District of New York.
The updated filing modifies one claim against Microsoft and drops a separate claim against OpenAI.
What the lawsuit actually argues #
The core allegation is straightforward: OpenAI and Microsoft used millions of Times articles without permission to train large language models, and those models now compete directly with the Times as an information source.
The newspaper is seeking billions in damages, citing lost subscription and advertising revenue as AI-powered tools increasingly serve as the first stop for information that readers might otherwise have paid to access.
An amended complaint filed in April 2024 added specific evidence to support those claims, including examples of GPT-4 outputs that reproduced Times content nearly verbatim and, in some cases, generated fabricated quotes attributed to real Times articles.
Microsoft has invested more than $10 billion in OpenAI and integrates its models directly into products like Copilot, which means Microsoft derives commercial benefit from the same training data at the center of the dispute.
Where the case stands #
As of April 2025, a judge had dismissed certain Digital Millennium Copyright Act claims against both OpenAI and Microsoft, but allowed the core copyright infringement allegations to proceed.
OpenAI’s primary defense has been fair use, the legal doctrine that allows limited use of copyrighted material under certain conditions, typically for commentary, education, or transformation. The argument is that training an AI model on text is transformative enough to qualify.
The Times has produced evidence that the outputs are not particularly transformative when a model can reproduce an article nearly word for word on request.
Why this matters beyond one newspaper’s balance sheet #
Large language models are trained on text. A significant portion of that text was produced by journalists, authors, academics, and creators who never consented to its use and received no compensation. The companies that built those models are now worth hundreds of billions of dollars.
If the Times prevails, it creates a legal template. Other media organizations would have standing to bring similar claims. Publishers, news agencies, and individual authors could argue for licensing fees retroactively or going forward.
If OpenAI and Microsoft prevail on fair use grounds, it effectively ratifies the current model: train on whatever is publicly accessible, build commercially valuable products, and let the content creators absorb the revenue loss.
For investors watching AI companies, a ruling that substantially expands copyright liability in AI training would hit every major model developer, not just OpenAI. Google, Meta, and others have trained models on similarly broad datasets. The Times case is the one with the most developed evidentiary record.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our