19 Newspapers. One Lawsuit. More Than $250 Million.
USA TODAY Co., the company formerly called Gannett, filed a copyright suit against OpenAI in Manhattan federal court on Thursday, October 8, over the journalism it says was used to train and feed ChatGPT.
Stacked newspapers, shown for illustration. Photo: psyberartist, CC BY 2.0, via Wikimedia Commons.
Another newsroom has had enough. On Thursday, October 8, USA TODAY Co., the publisher that used to be called Gannett, sued OpenAI in federal court in Manhattan. The docket on CourtListener lists it as USA Today Co., Inc. v. OpenAI Foundation, case 1:26-cv-08892, filed as a copyright infringement case with a jury demand. Reporting on the complaint puts the damages sought at more than $250 million.
Who is suing, and who is being sued
This isn't one paper. The complaint covers content from 19 publications owned by USA TODAY Co., including USA TODAY itself, The Tennessean, the Indy Star, the Detroit Free Press, The Detroit News, The Columbus Dispatch, The Oklahoman, The Arizona Republic, the Milwaukee Journal Sentinel and The Des Moines Register. The court record names a long list of plaintiff entities, among them GateHouse Media, the Detroit Newspaper Partnership and Gannett Satellite Information Network.
On the other side, the complaint names seven OpenAI entities: OpenAI Foundation, OpenAI GP, OAI International, OpenAI OpCo, OpenAI Global, OAI Corporation and OpenAI Group PBC.
What the papers say OpenAI did
According to Unite.AI's summary of the filing, the complaint says content from these publications makes up more than 160,000 entries in WebText, the corpus OpenAI built to train GPT 2, including 83,266 entries from usatoday.com and 12,994 from freep.com. It also says the papers' websites account for more than 122 million tokens in C4, a filtered snapshot of the Common Crawl web archive from 2019.
The part that hits ChatGPT users directly is about today's product, not old training runs. The filing includes examples where GPT 5.6 was asked to find and summarize a specific article by its title and returned what the complaint calls extensive summaries that paraphrase the original and follow its structure. The papers argue that this is the whole problem: if ChatGPT hands you the story, you have no reason to visit the paper or pay for a subscription. The complaint, as reported, quotes OpenAI's Head of ChatGPT saying that once ChatGPT gives an answer there is "no good reason to click" through to the source.
The money
The complaint leans on the statutory damages in U.S. copyright law: up to $150,000 for each work infringed willfully, plus up to $25,000 for each time copyright management information, the ownership details attached to an article, was stripped out. With hundreds of thousands of articles in play, it's easy to see how the papers get past $250 million. They're also asking the court to order OpenAI to stop.
A statement filed the same day asks that the case be treated as related to the consolidated OpenAI copyright litigation already pending in the same court, which includes the New York Times case. So this one is likely to join a very crowded docket. These are allegations, not findings, and at the time of the reports we read, OpenAI hadn't publicly responded.
Meanwhile, on the status page
The product kept wobbling while the lawyers filed. On Wednesday afternoon, October 7, OpenAI's status page logged "Elevated errors with GPT 5.6 Instant in FedRAMP workspaces," the government cloud version of ChatGPT, from about 4:42 to 5:03 PM Pacific. That night at about 8:17 PM Pacific, it opened "Some APAC users are seeing errors in ChatGPT Work, conversations, and GPTs," which wasn't marked recovered until about 9:51 PM. Earlier that day, "Dot in Codex may fail to create new threads" ran from about 9:31 to 11:06 AM Pacific.
Sources: CourtListener docket, USA Today Co., Inc. v. OpenAI Foundation; Unite.AI; Forbes; FedRAMP incident, October 7; APAC incident, October 7 and 8; Dot in Codex, October 7.