USA Today Co. and 13 affiliated entities sued OpenAI in federal court in New York, seeking more than $250 million. The filing also asks the court to order destruction of GPT models and training sets that use material from the 19 publications. These are requests and allegations, not a ruling.
The mechanism: the plaintiffs allege that articles were copied into training sources and that ChatGPT then reproduced or repackaged reporting for users. The mechanism matters because it links the provenance of data to what a deployed system returns. Why now: a complaint that joins alleged training use with alleged output substitution seeks remedies aimed at both the underlying materials and the resulting models. The filing reportedly identifies more than 160,000 entries in WebText and over 122 million tokens in a 2019 Common Crawl snapshot; Forbes also describes examples of GPT-5.6 producing detailed summaries of newspaper reporting.
The first implication: publishers are asking courts to consider remedies that could reach models and training sets, not only monetary compensation. That is a requested remedy, not evidence that such an order will issue or that the allegations are true. The evidence boundary: the supplied public account describes the complaint and the plaintiffs’ claims; it does not establish infringement, quantify proven harm, or report a court decision. The first watch condition: whether the court permits the case and requested model-destruction remedy to proceed, and how OpenAI responds to the allegations.