USA Today Co. sues OpenAI for $250 million over training on articles

USA Today Co. and several local newspapers in its portfolio have sued OpenAI. The publisher claims the company copied "hundreds of thousands" of articles to train its models, and it is asking for more than $250 million in damages. The complaint was filed in court on Thursday, October 8, 2026, and Reuters was the first to report it.

The lawsuit states that the unauthorized use of the content has caused "real and ongoing" harm to the group's publications. For a media company with a wide network of regional newsrooms, that is an important argument: local papers live on traffic and subscriptions, and their texts are what most often end up in training corpora.

Another lawsuit in a long queue

USA Today Co. is not the first. OpenAI is already being sued by The New York Times, whose complaint appeared in late 2023. Other plaintiffs include:

• The Intercept (together with Raw Story and AlterNet, lawsuits from 2024);
• Ziff Davis, owner of CNET, IGN and PCMag;
• Canada's CBC/Radio-Canada;
• Encyclopaedia Britannica.

The pattern is similar everywhere: the publisher believes its archives were used for training without a license and without payment, and that the result competes with the original. The new lawsuit adds American regional media to this picture.

What the numbers say

The published materials point to several key parameters of the case:

• amount sought: more than $250 million;
• volume of disputed content: "hundreds of thousands" of articles;
• defendant: OpenAI;
• plaintiffs: USA Today Co. and several local publications it owns;
• filing date: Thursday, October 8, 2026;
• basis: copyright infringement in training models.

The available material does not include OpenAI's position on this lawsuit. In similar disputes, the company has previously defended its right to train models on public data, citing fair use. The courts have not yet settled on a single line, and each new case adds material for future practice.

Why the amount matters

The $250 million demand sets a benchmark: the publisher values its damages at a sum comparable to the annual budgets of large newsrooms. That is also a signal to other content owners. If courts find such calculations well-founded, licensing archives for training will become a mandatory line item rather than an option. If not, model developers will keep room to train on the open web without separate agreements.

For now the outcome is open, and the number of cases keeps growing. For OpenAI, that means not only legal risk but also pressure in negotiations: companies with a lawsuit already filed bargain harder than those that have merely received a deal offer.

Practical takeaway

If you build a product on top of an LLM or train your own models, check where your data comes from: where the texts were taken, whether you hold licenses, and whether you can prove it with documents. Publishers, for their part, should record which of their materials are accessible to bots and preserve evidence of how their content is used. Watch the ruling in the New York Times case and OpenAI's response to the new lawsuit: they will most likely determine the terms on which the market buys and sells text for AI training.