USA TODAY Co., Inc. and its affiliated newspaper entities sued OpenAI in federal courtroom in New York on October 8, 2026, alleging the corporate skilled its GPT fashions on lots of of 1000’s of their copyrighted articles with out authorization and repackaged their journalism in chatbot outputs that substitute for the originals. The swimsuit seeks damages in extra of $250 million.
The Submitting and Requested Aid
The case, USA At this time Co., Inc. v. OpenAI Basis, was filed within the U.S. District Court docket for the Southern District of New York by lawyer Steven Lieberman of Rothwell, Figg, Ernst & Manbeck. The grievance was entered on October 8, 2026, accompanied by an exhibit of copyright registrations and a second exhibit of GPT-5.6 output examples, in keeping with the courtroom docket. Similar-day docket entries embrace a civil cowl sheet, a copyright discover type, a discover of look, and a request for issuance of summons.
An announcement of relatedness filed the identical day asks that the motion be handled as associated to the consolidated OpenAI copyright infringement litigation already pending in the identical courtroom. The grievance states that the plaintiffs be a part of a protracted record of copyright holders which have sued OpenAI and different AI corporations, lots of these instances consolidated in that courtroom.
Underneath the statutory framework the grievance cites, the plaintiffs could get better as much as $150,000 for every willful copyright infringement, plus as much as $25,000 per violation for the elimination of copyright administration data, the possession and rights information connected to revealed works. The plaintiffs allege OpenAI’s fashions copied lots of of 1000’s of articles and different supplies from their publications.
Coaching-Knowledge and Copying Allegations
In response to the grievance, OpenAI’s giant language fashions had been skilled on copyrighted materials scraped from the web with out authorization, no matter paywalls or different entry restrictions, and OpenAI used applications designed to strip away copyright administration data indicating the works had been protected. The grievance additionally cites written proof OpenAI submitted to a British Home of Lords inquiry in December 2023, during which the corporate said that as a result of copyright covers just about each form of human expression, limiting coaching information to public-domain works wouldn’t produce AI techniques that meet present wants.
The grievance states that content material from the plaintiffs’ publications includes greater than 160,000 entries in WebText, the inner corpus OpenAI constructed to coach GPT-2, together with 83,266 entries from usatoday.com and 12,994 entries from freep.com. It states the publications’ domains account for greater than 122 million tokens in C4, a filtered English-language subset of a 2019 snapshot of the Frequent Crawl net archive, and reproduces OpenAI’s revealed GPT-3 coaching combine, which weighted Frequent Crawl at 60 p.c and WebText2, an expanded model of the WebText corpus, at 22 p.c.
To argue that OpenAI understood what its fashions contained, the grievance quotes inner communications. It states that co-founder Greg Brockman advised colleagues the fashions had been notably good at predicting the textual content of reports articles, and {that a} 2020 presentation by then-research chief Dario Amodei listed information technology amongst GPT-3’s abilities. It quotes OpenAI’s VP of Analysis stating, “We practice our networks to memorize the coaching information — that’s their goal,” and cites June 2022 inner messages during which workers acknowledged that GPT-4 would have memorized a considerable amount of information and could be extremely efficient at regurgitating it.
The grievance additional alleges that over a three-year interval Microsoft offered OpenAI with a replica of the Bing Index, a compilation of billions of webpages scraped for Microsoft’s search engine that included the plaintiffs’ content material, below an initiative codenamed Challenge Taxi, and that Microsoft individually developed and operated a crawler referred to as Challenge Mango on OpenAI’s behalf, for which OpenAI paid Microsoft. It additionally alleges that OpenAI’s output filters didn’t suppress content material from any entity that had not sued the corporate, an strategy it says a Microsoft govt described internally as an unintentional cover-up.
GPT-5.6 Output Examples
The submitting reproduces examples during which GPT-5.6, prompted to seek out and summarize a selected article by title, produced what the grievance describes as in depth multi-section summaries paraphrasing the originals and following their structural group. The reproduced examples contain articles from the Indianapolis Star, the Detroit Free Press, The Knoxville Information-Sentinel, The Palm Seaside Submit, The Tennessean, The Enquirer, The Des Moines Register, The Courier-Journal, the Naples Each day Information, the Asbury Park Press, the Milwaukee Journal Sentinel, The Columbus Dispatch, The Oklahoman, the Star Information, the Democrat & Chronicle, The Bergen File, The Detroit Information, The Arizona Republic, and USA TODAY, with full variations connected as an exhibit.
The grievance alleges OpenAI affirmatively post-trained its fashions to summarize copyrighted articles in lieu of returning the articles themselves, producing substitutes that serve the identical informative objective because the originals. It quotes OpenAI’s Head of ChatGPT acknowledging that after ChatGPT provides a solution there’s “no good cause to click on” a hyperlink to the underlying supply, and cites an OpenAI engineer’s assertion that regardless of how prominently hyperlinks are displayed, customers won’t click on. When OpenAI’s outputs reproduce or repackage their content material, the plaintiffs allege, customers don’t have any cause to go to the unique sources or pay for a subscription.
The Events
The plaintiffs, all owned by USA TODAY Co., Inc., maintain copyrights in content material revealed by 19 publications, together with USA TODAY, The Tennessean, the Indy Star, The Bergen File, The Enquirer, the Asbury Park Press, the Democrat & Chronicle, The Knoxville Information-Sentinel, the Naples Each day Information, The Oklahoman, the Milwaukee Journal Sentinel, The Columbus Dispatch, The Arizona Republic, The Courier-Journal, The Des Moines Register, the Detroit Free Press, The Detroit Information, The Palm Seaside Submit, and the Star Information. The grievance describes USA TODAY Co., previously Gannett Co., Inc., as a diversified media firm tracing its historical past to 1906 that publishes USA TODAY and lots of of day by day publications, and states that its publications have gained dozens of Pulitzer Prizes and make use of lots of of journalists throughout 13 states.
The defendants are OpenAI Basis; OpenAI GP, LLC; OAI Worldwide, Inc.; OpenAI OpCo, LLC; OpenAI World, LLC; OAI Company; and OpenAI Group PBC, every described within the grievance as a San Francisco-based entity that was immediately concerned in or profited from the alleged infringement. The grievance recounts OpenAI’s October 28, 2025 recapitalization, below which the nonprofit grew to become the OpenAI Basis, holding fairness within the for-profit that OpenAI described as valued at roughly $130 billion, and the for-profit grew to become OpenAI Group PBC, a public profit company. Citing OpenAI’s personal disclosures, the grievance states that ChatGPT had greater than 900 million weekly lively customers and over 50 million paying subscribers as of March 2026.
The fashions at concern span GPT-1 via GPT-6.1 and GPT-OSS, together with all Immediate, Considering, mini, nano, and Professional variants, in keeping with the grievance. The plaintiffs demand a jury trial.
