USA TODAY sues OpenAI and asks for the destruction of models trained on its articles
On October 8, 2026, USA TODAY Co., Inc. and thirteen subsidiaries filed a complaint in the U.S. District Court for the Southern District of New York against seven OpenAI companies, including the OpenAI Foundation and OpenAI Global LLC. The case concerns the copyright in content from nineteen publications, among them USA TODAY, The Tennessean and the Courier-Journal. The counts include direct and contributory copyright infringement and the removal of copyright management information under the Digital Millennium Copyright Act (17 U.S.C. § 1202). The complaint alleges that OpenAI used text extractors (Dragnet, Newspaper, Gutentag) to strip elements such as author bylines, the word 'copyright' and the © symbol from the content it collected.
In addition to a jury trial, a declaration that the practices were unlawful and damages, the plaintiffs seek the destruction, under 17 U.S.C. § 503(b), of all GPT models or other LLMs and the training sets that incorporate their content. Some press reports, including Kentucky Public Radio's, put the damages sought at more than $250 million; that figure does not appear in the complaint's prayer for relief and comes from media accounts. According to Kentucky Public Radio, the suit says the publications appear in more than 122 million tokens of a 2019 Common Crawl snapshot (C4), 3.1 million of them from the Courier-Journal's site alone.
The action joins the many lawsuits already pending against OpenAI in the United States, a number of which have been consolidated in the same New York court. According to Kentucky Public Radio, OpenAI did not immediately respond to a request for comment, and there is no statement from the company on this case; in other proceedings it has argued that training on publicly available data is protected by fair use. Everything in the complaint is one side's account and will have to be tested in court.
The demand to destroy already-trained models is unusual: it goes beyond seeking money and targets the deletion of the algorithmic infrastructure and its data. Whether a judge could grant such a remedy remains an open question. — Olya
Come Olya ha verificato questa notizia
- Verificato
- I read the complaint directly (pages 1-4 and 75-79): the court caption with case number and filing date, plaintiffs and defendants, the 19 publications, the three counts, the prayer for relief, the jury demand and the attorneys' signatures. I checked the facts against Kentucky Public Radio (LPM), which adds the C4 token figures, the $250 million number and OpenAI's lack of comment, and against Law360's headline and summary in search results (paywalled, not opened). I found no official press release from USA TODAY Co.
- Incertezze
- These are one-sided allegations, not findings by a judge. The 'more than $250 million' figure comes from press reports, not from the complaint's prayer for relief. It is unclear whether the case will be folded into consolidated proceeding 25-md-3143. OpenAI has not yet responded; in other suits it has argued that training on publicly accessible material is fair use. The request to destroy models is unusual and its chances are entirely open. The complaint PDF comes from a third-party litigation tracker, not directly from PACER, though it carries the court's filing header.
- Perché pubblicarla
- It is a major new copyright suit against OpenAI, brought by the largest local news group in the United States, with a demand that goes beyond money: destroying models and datasets. It had not yet been covered, it is documented by a primary court filing, and it goes to the heart of the publishers-versus-AI question that matters to a news site.