USA Today Sues OpenAI for Over $250 Million in Copyright Case

USA Today Co. is seeking more than $250 million in damages from OpenAI, accusing the company of copying "hundreds of thousands" of its articles to train ChatGPT and related products. The publisher filed the copyright infringement lawsuit on Thursday in federal court, according to Reuters and a copy of the complaint reviewed by The Verge.

The complaint names USA Today Co. and several of the local newspapers it owns. It alleges OpenAI scraped and reproduced copyrighted journalism without permission or payment. The unauthorized use, the filing states, "has done real and continuing" harm to the publisher's outlets.

USA Today is the latest major news organization to take OpenAI to court over the construction of its training datasets. The publisher joins The New York Times, The Intercept, Ziff Davis, the parent company of CNET, CBC and Radio-Canada, and Encyclopaedia Britannica in litigation that tests whether the scraping of online journalism to build commercial AI products violates U.S. copyright law.

The legal question is narrow. It is whether the reproduction of copyrighted text during model training, even when the resulting model does not regurgitate that text, qualifies as fair use under the Copyright Act of 1976. The answer will determine whether publishers can license their archives for AI training or watch those archives become raw material for competing products without recourse.

What does the lawsuit allege?

The complaint targets the training pipeline itself, not specific ChatGPT outputs. According to the filing, OpenAI copied "hundreds of thousands" of articles published by USA Today and its local newspapers into the corpora used to build its language models. The publisher argues the copying was willful, systematic, and carried out for profit.

USA Today lawyers argue the scraping damaged the company three times over. It cost the publisher licensing revenue it could have charged for its archives. It let OpenAI build a commercial product that competes with the publisher's own news properties for audience attention. It allowed ChatGPT to surface news-style answers without paying the newsrooms that produced the underlying reporting.

The complaint seeks statutory damages, actual damages, and an injunction barring further alleged infringement. It does not name individual ChatGPT responses. It targets the act of copying.

Who else is suing OpenAI?

USA Today's filing joins a stack of copyright actions already pending against the AI company:

• The New York Times filed a copyright suit against OpenAI in 2023, alleging the company trained ChatGPT on millions of its articles.
• The Intercept sued in 2024 over the alleged use of its reporting without authorization.
• Ziff Davis, the parent company of CNET, filed against OpenAI over scraping of its technology publications.
• CBC and Radio-Canada filed a joint action, the first major copyright claim by a Canadian public broadcaster.
• Encyclopaedia Britannica filed a separate complaint alleging OpenAI ingested its reference content without a license.

The plaintiffs span daily newspapers, digital-native outlets, technology publishers, a public broadcaster, and an encyclopedic reference work — a near-complete sample of the media formats OpenAI's training data was built upon.

Why does the case carry weight?

USA Today is among the largest-circulation news organizations in the United States, and its owned local papers operate in markets across the country. A ruling against OpenAI in the case would extend across a wide footprint of U.S. media markets.

The $250 million damages request is also a benchmark. The demand exceeds the explicit damages totals sought by the smaller publishers in the OpenAI litigation stack. The figure signals that publishers are pricing AI training data at multiples of what open-license or fair-use assumptions had priced it before.

The damages demand also telegraphs strategy. A request north of $250 million suggests USA Today intends to prove willful infringement, the legal standard that unlocks statutory damages up to $150,000 per work under federal copyright law. Multiplied across hundreds of thousands of articles, the theoretical exposure for OpenAI runs into the tens of billions.

A win for USA Today would give publishers a template for proving infringement, damages, and willfulness. A loss, or a narrow fair-use ruling, would harden the legal ground under OpenAI's training practices and weaken the leverage of every other plaintiff in the queue.

What is OpenAI's position?

OpenAI has consistently argued that training large language models on publicly available web material constitutes fair use under U.S. copyright law. The company has emphasized that its models transform source content rather than reproduce it verbatim, and that ChatGPT includes citations and links when surfacing news material.

In court filings in related cases, OpenAI has pointed to rulings in favor of search engines and thumbnail image services as precedent. The company has framed its training practices as analogous to Google's indexing of the web — a transformative use that requires no per-article license.

OpenAI has also pointed to its licensing track as evidence that it negotiates with publishers when asked. The company has framed the litigation as a disagreement over contract terms rather than over the legality of training itself.

USA Today's complaint does not engage with the licensing argument. It alleges that OpenAI never sought a license before copying.

What comes next?

The case will move through initial motions on jurisdiction and service. A federal judge will likely consider merging it with related OpenAI filings under multidistrict procedures. Discovery could last years and force OpenAI to disclose its training data composition and the internal communications that shaped its data strategy.

For OpenAI, the litigation is no longer a single case. It is an industrial-scale legal operation layered on top of the company's product, infrastructure, and policy work. The publisher side, by contrast, is fighting for a share of the commercial value their archives created.

The legal fight will determine whether the training of foundation models on journalism is treated as fair use, as OpenAI argues, or as a compensable taking, as the publishers contend. A federal court, and possibly the Supreme Court, will settle the question.

The articles were the feedstock. The publisher wants paid.