USA Today Sues OpenAI for $250 Million Over AI Training Data

USA Today Co. and several local newspapers it owns have sued OpenAI, alleging the company copied "hundreds of thousands" of their articles to train its AI models and seeking damages of more than $250 million. The filing was made on Thursday, October 8, 2026, according to The Verge, which cites Reuters as the first to report it. OpenAI had not responded to The Verge's request for comment when the story ran.

It is the newest entry in a long line of publisher suits against OpenAI, and it brings a group of local newspapers into the list. For people who build with or study these models, the case is another data point in an unresolved question: what does it cost, legally, to train a frontier model on the open web's journalism?

What the USA Today lawsuit says

The complaint's central line, as quoted by The Verge, is blunt: "OpenAI's commercial success rests on large-scale copyright infringement." It further alleges that OpenAI never asked permission and instead "took" the publisher's content to "build products worth hundreds of billions of dollars." The publisher also says the unauthorized use "has done real and continuing" harm to its outlets.

We have only the press accounts of the filing, not the complaint itself, so details such as the exact causes of action, the specific models named, or how the $250 million figure was calculated are not something this post can confirm. Where reporting is silent, we say so rather than guess.

What people are asking about the case

Can a newspaper really prove its articles were used for training? That is usually the hard part. Companies rarely publish full training-set lists, so plaintiffs rely on circumstantial evidence: models that reproduce distinctive passages, crawler logs showing repeated access, and statements from the developer about data sources. Discovery, the phase where each side must hand over documents, is where these cases often turn. Expect the early fights to be about what OpenAI must disclose regarding its training data.

Why sue now rather than negotiate? Publishers can do both. Filing a suit sets a price anchor, creates press attention, and starts the clock on discovery. It also signals to other AI companies that unlicensed use carries a risk. Whether USA Today Co. is also in licensing talks with anyone is not something the reporting tells us.

Does it matter that the plaintiffs include local papers? Yes, in two ways. Naming individual papers like The Tennessean and The Oklahoman lets the plaintiffs describe harm to specific communities rather than an abstract industry.

Is there a risk to users of ChatGPT? The suit asks for damages and, as reported, does not target individual users. Any change that touches users would come from a court order or a settlement, and neither exists yet.

Who else is suing OpenAI

The Verge's report lists the existing plaintiffs, and the list is long enough to be its own story:

• The New York Times
• The Intercept
• Ziff Davis, owner of CNET
• CBC/Radio-Canada
• Encyclopaedia Britannica and Merriam-Webster
• The Seattle Times (see also the Verge's related report on The Seattle Times and Newsday suing OpenAI and Microsoft)
• A coalition of nearly 400 local newspapers

The Verge also points to a related development: the Trump administration is supporting OpenAI in the New York Times copyright case. That matters for context, because it shows the federal government taking a side in at least one of these fights, at the same moment a new plaintiff files.

Notice the mix. Early suits came from national brands with deep legal budgets. The recent wave includes reference publishers, regional papers, and now USA Today Co. with its local papers. Each new plaintiff makes the "nobody was harmed" defense harder to sustain in the court of public opinion, even though the courts will decide on the law.

Why publishers keep suing: the economics

Local and national news organizations have watched their search traffic erode as AI answers absorb queries that used to send readers to their pages. We covered the traffic side in our post on publishers threatening to block Google's crawlers after a 40 percent traffic drop. A lawsuit is the other lever: if you cannot stop the crawl, you can try to price it.

There is also a licensing market emerging underneath the litigation. OpenAI has signed content deals with some publishers, and image licensing has followed the same path, as we traced in the Getty and OpenAI deal history. A lawsuit can be read as a negotiating position as much as a demand for a verdict: publishers who sue often end up at the table. We cannot say whether that is the intent here, and the filing as reported does not say.

The sums are also telling. A demand north of $250 million for one publisher group suggests plaintiffs are now pricing training data in the same range as major enterprise contracts, not as a rounding error.

The legal question underneath

Almost all of these cases turn on the same two questions:

1. Is copying text to train a model a "fair use"? Defendants argue training is transformative: the model learns patterns rather than republishing articles. Plaintiffs argue it is a commercial substitute that competes with the original.
2. Does the model output reproduce or replace the source? Plaintiffs point to cases where a chatbot regurgitates paragraphs or summarizes paywalled reporting. Defendants point to safeguards and to the fact that most outputs are novel text.

Neither question has a settled, universally applicable answer. Courts have reached different conclusions depending on the facts, the kind of work, and how the training data was obtained. That is why we advise treating any one filing, including this one, as an input rather than a verdict. For a broader legal grounding, our guide to AI and law covers how these concepts work in practice.

It is also not just OpenAI. Music publishers are pursuing Anthropic over song lyrics, with demands that scale per work; see why Sony and Warner want $150,000 per song from Anthropic. The per-work damages math is the reason "hundreds of thousands of articles" becomes a nine-figure number so quickly: statutory damages in US copyright law are calculated per infringed work.

What this means for people building with these models

If you are a developer, an agency, or a team shipping products on top of OpenAI's API, nothing about your access changed today. But there are four second-order effects worth tracking:

• Pricing pressure. Lawsuits and licensing deals are costs. If they stack up, they can show up in API prices, tier changes, or usage limits. OpenAI's recent pricing and speed-tier moves, like the Ultrafast tier covered after DevDay, already show how fast the cost structure shifts.
• Output behavior. Courts and settlements can force guardrails around reproducing news content, which affects summarization and retrieval apps.
• Data provenance for your own fine-tunes. If you scrape news for training or retrieval, the same questions apply to you at smaller scale. Keep records of sources and licenses.
• Vendor risk. Several leading labs now face copyright litigation from different industries. Diversifying across providers is a reasonable hedge, not a prediction about who wins.

Context: a company under pressure from several sides

The suit lands in a week of heavy scrutiny of OpenAI. The company recently drew attention over the firing of three safety researchers, and legal pressure now comes from publishers and other plaintiffs. For a wider view of the litigation climate across labs, see our look at the lawsuit accusing Anthropic, OpenAI, Google and SpaceXAI of coordinating a slowdown. For the longer arc of how we got here, our history of artificial intelligence places data-hungry scaling in context.

What to watch next

1. OpenAI's response. An answer, a motion to dismiss, or a statement will show whether the company contests fair use or the specifics of the copying.
2. Consolidation. Courts sometimes coordinate similar cases. Watch for moves that bring the publisher suits together.
3. A settlement or license. Large plaintiffs often convert litigation into a licensing deal. A deal would tell us how the market prices news archives.
4. The New York Times case. It is the bellwether. Rulings there will shape how everyone else argues.
5. The administration's stance. Federal support for OpenAI in one case could influence the policy backdrop for the rest.

Bottom line

USA Today Co. wants more than $250 million and a court finding that OpenAI's commercial success rests on copying journalism without permission. OpenAI has not yet answered. The case adds weight to a pattern, rather than resolving it, and the practical advice for builders is steady: assume rules around training data will tighten, document your own data sources, and avoid single-provider lock-in.

Reporting in this post reflects press accounts as of October 8, 2026. We have not reviewed the complaint itself, and details may change as OpenAI responds.

Related reading

• Why Sony and Warner want $150,000 per song from Anthropic
• Getty and OpenAI: image copyright deal history
• Google Zero: publishers threaten to block crawlers
• The cartel lawsuit against Anthropic, OpenAI, Google and SpaceXAI
• AI and law: a practical guide
• OpenAI fires three safety researchers
• History of artificial intelligence, 1950 to 2026