A newspaper's suit tested reproduction, not just training

The writer's problem

The New York Times faced a different kind of exposure than book authors: not only training on its journalism, but chatbots allegedly capable of reciting large blocks of its articles back to users. On 27 December 2023 the Times sued OpenAI and Microsoft in the Southern District of New York, filing a complaint that, according to the paper's own report of its filing, sought unspecified damages and asked the companies to destroy any models and training data built from its work.

What the documents show

The complaint alleges millions of Times articles were used to train GPT-based models, and separately alleges the resulting chatbots can reproduce close paraphrases or near-verbatim passages of Times journalism, a claim distinct from the pure training-data theory in the book-author cases. The paper's own coverage states it had approached Microsoft and OpenAI in April 2023 seeking a licensing arrangement before litigation. OpenAI responded in its own post, OpenAI and journalism, published 8 January 2024, stating it viewed the lawsuit as without merit, that regurgitation of training text is a rare bug it works to reduce, and that the examples the Times found came from prompts it says were engineered to induce the behavior. That is OpenAI's own account of events and its own characterization of the evidence; it is not a court finding, and the complaint's allegations about reproduction likewise remain allegations, not verified facts, until a court rules on them.

The editorial choice

Reporting on this case should keep the two theories separate: a training-data claim shared with the book-author suits, and a reproduction claim specific to how the outputs are used, since a ruling on one would not automatically resolve the other. This case was later folded into the same consolidated proceeding as the Authors Guild's suit, under the judicial panel's 2025 transfer order, without merging its distinct legal theories.

What stays with the author

The Times's own reporting, and its journalists' bylines, remain its own regardless of the suit's outcome; the dispute is about a chatbot's output, not about who wrote the underlying articles. What the sources do not resolve is whether a court will treat verbatim-style outputs differently from a general training-data claim.

• Does a report on this case describe the training claim, the reproduction claim, or conflate the two?
• Is a quoted account of what a chatbot produced coming from the complaint, or from OpenAI's response to it?
• Has this case been decided, or is it still consolidated with other pending suits?

The Times case adds a reproduction theory that the book-author suits do not raise in the same way, and the two should not be read as interchangeable versions of the same claim.