USA Today's Owner Sues OpenAI, Seeking More Than $250 Million Over AI Training

USA TODAY Co. -- the company formerly named Gannett -- sued OpenAI on Oct. 8 in the U.S. District Court for the Southern District of New York, docketed as No. 1:26-cv-08892. The complaint names USA TODAY itself plus affiliated papers including the Detroit Free Press, The Arizona Republic, The Tennessean, The Indianapolis Star, The Columbus Dispatch and The Milwaukee Journal Sentinel, and alleges OpenAI trained ChatGPT on hundreds of thousands of their articles without permission or payment.

The complaint's specific claim is narrower than "OpenAI read our website": it alleges OpenAI's own WebText training corpus contains more than 160,000 entries sourced from USA TODAY Co. domains, 83,266 of them from usatoday.com alone, and that GPT-5.6's outputs mirror the structure and content of the original articles closely enough to constitute infringement rather than independent generation. Beyond the $250 million in damages, the publishers want a court order blocking further use of their work and, per some reporting on the filing, the destruction of any model trained on it -- the same remedy book publishers and The New York Times have separately sought against OpenAI in the same court's multidistrict litigation. OpenAI had not filed a public response as of this writing.

This is the latest, not the first, publisher to make this exact argument against OpenAI in this exact courthouse. The New York Times' own case against OpenAI and Microsoft is already well into discovery in the Southern District of New York, and a separate group of newspapers owned by Alden Global Capital filed a comparable suit in 2024. Whether this case proceeds on its own or gets consolidated with the Times' litigation is the first procedural question a judge will have to answer -- long before anyone reaches whether the underlying copying claim holds up.