USA Today Co. Sues OpenAI for More Than $250 Million, Citing 19 Publications and a Project Named 'Mango'
The complaint alleges training on hundreds of thousands of articles and summaries that substitute for the originals; the court has ruled on none of it.
This one is a lawsuit, so we are going to keep it to what has been alleged, by whom, and where. The court has not ruled on any of it.
On October 8, USA Today Co., the company formerly known as Gannett, sued OpenAI in federal court in the Southern District of New York. The case is captioned USA Today Co., Inc. v. OpenAI Foundation, per Unite.AI’s reading of the complaint, and seeks more than $250 million in damages, along with a court order blocking the alleged infringement, per Runtime Wire.
OpenAI had not immediately commented in the coverage we read.
The allegations, as reported. The plaintiffs say OpenAI trained its GPT models on hundreds of thousands of their copyrighted articles without authorization. They also allege that ChatGPT’s outputs repackage their journalism in ways that substitute for the originals. The complaint covers content from 19 publications, including USA Today, The Tennessean and the Detroit Free Press, with the Verge-sourced Crypto Briefing story naming IndyStar, The Columbus Dispatch and The Oklahoman as well.
Unite.AI’s detailed summary puts specific numbers on the training-data claims: more than 160,000 entries from the plaintiffs’ publications in WebText, of which 83,266 are from usatoday.com and 12,994 from freep.com, and more than 122 million tokens in C4. These are allegations about datasets, repeated here as allegations.
For statutory exposure, the complaint cites up to $150,000 per willful infringement, plus up to $25,000 per violation for removing copyright management information. At those figures, “more than $250 million” starts to read as a modest starting position, the way a first offer on a car does.
The code names. The complaint, as summarized by Unite.AI, also alleges Microsoft gave OpenAI a copy of the Bing Index under a project called “Project Taxi,” and operated a crawler called “Project Mango” on OpenAI’s behalf, which OpenAI paid for. These are allegations. No finding has been made that either project did what the complaint says.
We can, however, note that the plaintiffs have given the world a litigation drama with a taxi and a mango. In a field that normally names its products after Greek letters and the color of the loading screen, this is the first time a crawler has sounded like a smoothie.
The output examples. The complaint reproduces what it says are GPT-5.6 summaries of articles from 19 publications, and alleges OpenAI post-trained its models to summarize articles instead of returning them. That, in the plaintiffs’ framing, is the heart of the market-substitution claim: if the chatbot gives you the article’s gist, the article’s own page loses the visit.
The complaint also lists models from GPT-1 through GPT-6.1 and GPT-OSS, and cites OpenAI’s reported 900 million weekly ChatGPT users and over 50 million paying subscribers as of March 2026.
For readers keeping track of the broader docket: the suit joins existing copyright cases involving The New York Times, The Intercept, Ziff Davis, CBC/Radio-Canada, Encyclopaedia Britannica, Merriam-Webster and other publishers, per Crypto Briefing. Per Unite.AI, a statement of relatedness asks that the new case be treated as related to the consolidated OpenAI copyright litigation already pending in the same court. The plaintiffs have demanded a jury trial.
What happens now. Mostly, paperwork. A defendant gets time to respond. A judge decides whether the new case belongs alongside the consolidated ones. Motions are filed. Documents are exchanged. Eventually, perhaps years from now, someone reads a very long sentence aloud in a courtroom containing the word “tokenization.”
The substantive questions are the ones this litigation has already been circling: whether training on copyrighted articles is permitted use, and whether the outputs compete with the originals. They are, as the saying goes, questions presented. We don’t have the answers, and neither, as far as this reporting shows, does the docket.
Meanwhile the strongest line in the coverage may be the quietest one. Unite.AI notes these are allegations from the complaint, not findings of the court. That is not a hedge. It is the whole shape of the thing. A complaint is a story told by one side, with exhibits, and the exhibits here, if they’re accurate, include a bot summarizing a local newspaper’s reporting back to the person who would otherwise have clicked on the local newspaper.
A word on the genre. News organizations suing AI developers have become a recurring beat, and the industry that builds the summarizers is being sued by the industry that gets summarized, and a site like this one sits awkwardly in the middle, summarizing the summarizers.
We make no claim about any party’s training data. We are, for the record, not suing anyone.
Come back when there’s a ruling. Until then, the score is: allegations, 19 publications, one mango, and a damages figure that begins with a “more than.”
Sources: