Gannett Sues OpenAI for $250 Million Over 19 Newspapers' Journalism
Gannett, USA Today's parent company, sued OpenAI in the Southern District of New York for over $250 million, alleging willful copyright infringement of reporting from 19 newspapers, including copying for training and removal of copyright notices. The complaint also asks for destruction of models built on the material, making it one of the most aggressive publisher suits to date.
On this page
The suit, and the number
Gannett, the parent of USA Today and the largest newspaper chain in the United States, sued OpenAI in the Southern District of New York, with USA Today Co. and 13 affiliated entities joining, seeking more than $250 million in damages for willful copyright infringement of reporting from 19 newspapers. Per-work statutory damages of up to $150,000 are on the table, and the complaint alleges OpenAI copied the journalism for training, reproduced it in outputs, and stripped copyright management information from what its models ingested.
The complaint, which is public, goes further than damages: it asks for destruction of models built on the protected material. That remedy, if a court ever took it seriously, would be unprecedented in the generative-AI cases filed so far, and it signals Gannett is negotiating with leverage rather than settling for a licensing check.
What makes this suit different from the NYT's
The New York Times' December 2023 suit against OpenAI was the landmark; Gannett's is the mass-market escalation. Where the Times argued degradation of its premium journalism, Gannett speaks for local reporting across 19 papers, the category least able to absorb litigation costs and most dependent on search and referral traffic that AI answers displace. The memorization allegation is also direct: the complaint claims OpenAI's models copied and memorized the journalism, then filtered copyright notices from what was ingested.
OpenAI's litigation docket now includes the Times, Gannett, the Authors Guild lineup, and news organizations on several continents, while the company simultaneously signs licensing deals with other publishers. The split strategy, license some, litigate the rest, is reaching the point where a court will eventually have to say what training on news actually is.
Why the destruction remedy matters to everyone else
For people who run local models, the uncomfortable wrinkle is that remedies aimed at models are remedies aimed at weights. A destruction order against one commercial model would not touch open-weight releases directly, but it would set precedent for what courts can order when training data is adjudicated infringing, and open-weight publishers would feel that chill in their next release decisions.
The realistic outcomes remain licensing, damages, or settlement, and Gannett's $250 million opening number is an anchor rather than an expectation. What the suit changes is the price of the status quo: every day of training on unlicensed news is now attached to a specific complaint, a specific newspaper count, and a specific demand that the model itself be destroyed. OpenAI has not yet responded publicly to this filing, and its answer will say a lot about which publishers get licensed and which get litigated.