USA TODAY sues OpenAI over 19 publications and ChatGPT outputs

A new publisher case, with a broader target

USA TODAY Co. and 13 affiliated entities sued OpenAI in the U.S. District Court for the Southern District of New York on October 8, 2026, alleging that OpenAI copied reporting from 19 publications without authorization and used it in connection with GPT-based products including ChatGPT [1]. The complaint names USA TODAY, The Tennessean, Indy Star, The Bergen Record, The Enquirer, Asbury Park Press, Democrat & Chronicle, The Knoxville News-Sentinel, Naples Daily News, The Oklahoman, Milwaukee Journal Sentinel, The Columbus Dispatch, The Arizona Republic, The Courier-Journal, The Des Moines Register, Detroit Free Press, The Detroit News, The Palm Beach Post and Star News [1].

The suit seeks damages “in excess of $250 million,” along with injunctive relief and the destruction, under copyright law, of GPT or other large-language-model systems and training sets that allegedly incorporate the publishers’ content [1]. That request is important because it is not just a demand for money; it asks the court to restrict how OpenAI can continue using disputed material and to address systems already built with it [1].

OpenAI has not yet been found liable. The filing is a complaint, not a judgment, and its most explosive statements are allegations by the plaintiffs. The Verge reported that OpenAI did not immediately respond to its request for comment, while Bloomberg Law framed the case as another expansion of the copyright fight over OpenAI’s use of news content [2][3].

What USA TODAY says OpenAI did

At the center of the lawsuit is a claim that OpenAI used copyrighted journalism in several stages: acquiring or scraping copies of the publishers’ websites, training and fine-tuning models, and operating products that allegedly reproduce or repackage articles for users [1]. The complaint says OpenAI’s models were trained on copyrighted material scraped from the web without authorization, including material behind access restrictions, and that copyright-management information such as titles, bylines and notices was allegedly stripped away [1].

The publishers also try to quantify the presence of their work in known datasets. The complaint alleges that USA TODAY publications account for more than 160,000 entries in WebText, including 83,266 entries from usatoday.com, and that their domains represent more than 122 million tokens in C4, a filtered Common Crawl subset [1]. Those figures matter because AI copyright cases often turn on proof: plaintiffs must connect their works to training data, outputs or both.

The claims are pleaded under three main theories: direct copyright infringement, vicarious copyright infringement and removal of copyright management information under the Digital Millennium Copyright Act [1]. USA TODAY also argues that statutory damages could reach up to $150,000 for each willful copyright infringement and up to $25,000 per violation for removal of copyright-management information [1].

Why this case is not only about training

The most consequential part of the case may be its focus on current product behavior. Many publisher lawsuits against AI companies emphasize training: whether copying books, articles or web pages into datasets is protected by fair use. USA TODAY’s complaint goes further by targeting the operation of AI products, including search-like and retrieval-augmented answers that allegedly summarize specific articles so fully that users no longer need to click through [1].

The complaint describes retrieval-augmented generation, or RAG, as a process in which a model receives a query, retrieves relevant documents, combines them with the user’s prompt, and generates a natural-language response [1]. The publishers argue that this can produce answers that closely paraphrase or extensively copy works, even when the model itself may not have memorized the article [1]. In practical terms, USA TODAY is saying the harm is not frozen in the original act of training; it allegedly continues every time a product turns journalism into a substitute answer.

That distinction could matter for remedies. If a court sees the dispute only as a past training-data issue, damages and future dataset governance may dominate. If the court accepts that outputs and summaries are independent acts of infringement, the case could also affect product design, search interfaces, source linking, output filters and licensing terms for live news retrieval.

The GPT-5.6 examples

USA TODAY’s complaint includes examples involving GPT-5.6 prompts that ask the system to find and summarize named articles by title [1]. The filing alleges that GPT-5.6 generated extensive summaries of articles from outlets including Indy Star, Detroit Free Press, The Knoxville News-Sentinel, The Palm Beach Post, The Tennessean, The Enquirer, The Des Moines Register, The Courier-Journal, Naples Daily News, Asbury Park Press, Milwaukee Journal Sentinel, The Columbus Dispatch, The Oklahoman, Democrat & Chronicle, The Bergen Record, The Detroit News, USA TODAY and The Arizona Republic [1].

The complaint characterizes those outputs as multi-section summaries that preserve structure, sequence or substantive content from the original stories [1]. For the publishers, that is the legal bridge between “training” and “operation”: if a user can ask for an in-depth summary of a named article and receive a detailed substitute, the alleged injury is traffic loss, subscription loss and weakened licensing value [1].

The Verge’s summary highlighted the same demand for more than $250 million and quoted the complaint’s claim that OpenAI’s unauthorized use caused “real and continuing” harm to the outlets [2]. Bloomberg Law likewise reported that the plaintiffs own copyrights in content across 19 publications and that the complaint says OpenAI targeted news content [3].

Terms of service, robots.txt and licensing pressure

USA TODAY also points to its own restrictions. The complaint says the publications’ terms bar using site content to develop or improve AI systems, including training, fine-tuning, grounding and retrieval-augmented generation, and says the publications block OpenAI crawlers through robots.txt [1]. The filing further alleges that detected AI crawlers are redirected to a page stating that crawling and scraping are not permitted [1].

That argument is designed to undercut any claim that OpenAI reasonably believed the use was permitted. It also reflects a wider industry shift: publishers are turning technical signals, site terms and licensing history into evidence. The complaint notes that OpenAI has licensing agreements with more than a dozen news and media organizations, including the Associated Press, Axel Springer, The Atlantic and Vox Media, and argues that those deals show OpenAI understands journalism has commercial value and requires permission [1].

For OpenAI and other AI developers, that is the provenance problem in sharper form. It is no longer enough to say that training data came from the open web. Courts, publishers and enterprise customers increasingly want to know which domains were used, under what terms, whether content was behind paywalls, whether robots.txt exclusions were respected, and whether outputs can reproduce or replace the original work.

What happens next

Procedurally, the case begins as another Southern District of New York copyright action against OpenAI. Unite.AI reported that the docket included the complaint, copyright registration materials and GPT-5.6 output examples, and that a statement of relatedness asked for the action to be treated as related to consolidated OpenAI copyright litigation already pending in the same court [4].

The next milestones are likely to include service, appearances by OpenAI’s counsel, and either an answer or a motion challenging the complaint. The central legal fight will likely involve fair use, causation, proof of copying, the legal status of summaries, and whether removal of copyright-management information can be shown at scale.

For publishers, the lawsuit is another attempt to convert archives and current reporting into licensing leverage. For AI companies, it raises the cost of ambiguity: if news content appears in training sets and products can produce article-level substitutes, “we scraped the web” may no longer be a sufficient factual answer. The case is still only at the allegation stage, but its theory is clear: the complaint treats journalism not merely as historical training fuel, but as live product infrastructure that publishers say OpenAI used without paying.

Sources from the last 72 hours

1. [1]Complaint, USA Today Co., Inc. v. OpenAI Foundation, No. 1:26-cv-08892 (S.D.N.Y.)Oct 8, 2026, 2:00 AM
2. [2]USA Today becomes the latest publisher to sue OpenAIOct 8, 2026, 7:58 PM
3. [3]USA Today, News Outlets Join OpenAI Copyright Infringement FightOct 8, 2026, 7:59 PM
4. [4]USA TODAY Sues OpenAI Over Copyrighted News Content in AI TrainingOct 8, 2026, 2:00 AM

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.