USA Today sues OpenAI for copyright infringement over AI training

USA Today Co. and 18 of its affiliated newspapers have taken OpenAI to federal court. They allege the ChatGPT maker trained its AI models on their journalism without asking or paying.

The suit was filed on October 8, 2026, in the US District Court for the Southern District of New York. It seeks more than $250 million in damages. The plaintiffs also want the destruction of any models trained on their copyrighted work.

What the lawsuit alleges

The complaint, filed as Case No. 1:26-cv-08892, claims OpenAI used hundreds of thousands of articles from the plaintiffs’ publications to build its AI systems. According to the filing, none of that material was licensed.

The publishers back the claim with specific numbers. The complaint says more than 160,000 entries from the plaintiffs’ domains appear in OpenAI’s WebText corpus, a collection of web text used to train language models.

Of those, 83,266 allegedly came from usatoday.com alone. The rest are spread across the affiliated papers.

The suit goes further than the training stage. It points to the capabilities of GPT-5.6, alleging that the model’s outputs mirror both the structure and the content of the original articles.

The legal claims and the math

The plaintiffs are pursuing several theories at once. These include direct copyright infringement, vicarious copyright infringement, and violations tied to the removal of copyright management information, or CMI.

CMI is the identifying data attached to a work, such as bylines, titles, and copyright notices. The complaint alleges OpenAI stripped that information from the articles it copied. Removing it falls under the Digital Millennium Copyright Act.

Statutory damages could potentially reach up to $150,000 per willful infringement, according to the filing’s framing. The publishers also seek $25,000 per DMCA violation involving CMI removal.

The plaintiffs are also asking for injunctive relief, which would bar OpenAI from continuing the alleged conduct.

Steven Lieberman of the law firm Rothwell, Figg, Ernst & Manbeck represents the plaintiffs. OpenAI had not responded to the lawsuit or commented publicly as of October 9, 2026.

A crowded courtroom

USA Today is not first in line. The New York Times has already brought its own case against OpenAI, as has The Intercept. A coalition of nearly 400 local papers has filed similar claims too.

What this means

The specificity of this complaint stands out. Naming the WebText corpus and attaching a count of 83,266 usatoday.com entries gives the plaintiffs a concrete foundation to argue from, rather than asking a court to infer what went into training.

The CMI claims could prove to be the sleeper issue. Copyright infringement arguments over AI training remain contested, but stripping bylines and notices is a narrower allegation with its own statutory penalty attached.