Six Publishers Sue Microsoft and OpenAI Over ChatGPT and Copilot Training Data

The filing at a glance​

The case is Emmerich Newspapers, Inc. et al. v. Microsoft Corp. et al., case number 3:26-cv-00791, in the U.S. District Court for the Southern District of Mississippi. Six publishing companies led by Emmerich Newspapers filed against Microsoft and ten OpenAI entities. The complaint is stamped "Filed 10/02/26". CourtListener lists the case as assigned to Judge Henry Travillion Wingate.

The plaintiffs are Emmerich Newspapers, Ojai Media, Coopwood Publishing Group, Coopwood Magazine, Coopwood Media Group and Coopwood Newspapers. The complaint calls Emmerich one of the largest privately owned newspaper chains in Mississippi, with additional papers in Louisiana and Arkansas. According to FourWeekMBA's reading of the complaint, Ojai owns the Ojai Valley News in California, and the Coopwood companies publish titles such as the Mississippi Business Journal and Delta Magazine.

The defendants are more numerous than the shorthand "Microsoft and OpenAI" suggests. The caption names Microsoft Corporation and ten OpenAI entities, which the complaint calls a "web of interrelated entities."

What the publishers allege​

Neowin's report and FourWeekMBA's account of the complaint agree on the main points:

• Scraping. The publishers say the defendants crawled their sites and copied tens of thousands of articles. They say this included content behind paywalls, and that it happened without permission or payment. FourWeekMBA's account says the complaint claims the copying was repeated as models were updated.
• Datasets. The complaint cites OpenAI's own descriptions of training data. For GPT-2 that is WebText, built from links shared on Reddit. For GPT-3 it names Common Crawl, WebText2, Books1, Books2 and Wikipedia. It alleges that analyses of Common Crawl and OpenWebText found the plaintiffs' material. The complaint does not say how those analyses were done, and it does not list the articles found.
• Stripped attribution. The plaintiffs say their pages carried author credits, publication names, copyright notices and terms of use. They allege these were removed before training. The complaint points to the Dragnet and Newspaper content extractors OpenAI described using for WebText. It alleges, on information and belief, that OpenAI used them to remove this information intentionally. Intent is unproven.
• Outputs. The complaint relies mainly on testing described in other publishers' lawsuits, such as those from The New York Times and the New York Daily News. It alleges, on information and belief, that earlier ChatGPT versions would also produce near-verbatim copies of the plaintiffs' work. It sets out no example output from a plaintiff's article.
• Retrieval features. The complaint also says features like ChatGPT Search and Deep Research pull content from publishers' sites and fold it into answers.

Where Microsoft fits​

Microsoft's role is why this matters to Windows and Microsoft 365 readers. The complaint's three counts do not apply to every defendant in the same way.

On Count I, FourWeekMBA's account says the complaint ties Microsoft to storing and processing training data on its supercomputing platform. It also links Microsoft to distributing output through Copilot. The vicarious count says Microsoft controlled and profited from OpenAI's conduct and could have stopped it. The complaint alleges Microsoft built a dedicated supercomputer for OpenAI and, on information and belief, was involved in every step of training.

The stripped-attribution count is aimed at OpenAI, not Microsoft. Headlines that say both companies stripped the information flatten that distinction.

What the publishers want​

The plaintiffs seek statutory damages, including for alleged willful infringement. They also seek compensatory damages, restitution, disgorgement, costs and attorney's fees, and they demand a jury trial. No dollar figure appears in the prayer for relief.

The most unusual request is an injunction. The complaint asks for an order under 17 U.S.C. § 503(b) requiring the defendants to remove all copies of the registered works from all GPT or other LLM models and training sets. The complaint offers no method for pulling specific articles out of a trained model. Researchers call that problem "unlearning," and it is hard. Courts and the defendants will probably have a lot to say about whether such an order is technically feasible. That is analysis, not something the complaint addresses.

A growing pile of local-news suits​

This is not the first local-news suit against the pair, and Neowin notes that The New York Times sued earlier. Other cases the search turned up:

• Eight Alden-owned daily newspapers sued OpenAI and Microsoft in 2024. They included the New York Daily News and the Chicago Tribune.
• Courthouse News reported a suit in Manhattan from a group of publishers owning nearly 400 newspapers. That group accused both companies of scraping content to train ChatGPT and Microsoft Copilot. Shacknews put Bloomberg Law's report of that suit on June 24, 2026.
• Separately, Microsoft was hit with another copyright-focused shareholder derivative suit, Subramanian v. Microsoft, reported the same week.

The Mississippi case sets itself apart through the plaintiffs and the venue. It is a small group of independent publishers, and it is in a different federal district from the New York cases. Whether it stays in Mississippi or gets consolidated or transferred is unknown. Nothing in the material I reviewed addresses that.

What is not known​

• No ruling. FourWeekMBA's reading of the docket, last updated October 5, showed only the complaint, an issued summons and a copyright notice.
• No response located. Neowin says Microsoft was contacted for comment. I found no statement from Microsoft or OpenAI about this complaint.
• No independent verification. FourWeekMBA says it read the complaint and docket but verified none of the allegations. I could not read the complaint directly. Details here come from that account and from Neowin.
• Unchecked figures. The complaint also cites OpenAI valuation and revenue figures. I have left those out because they are unverified.

What it means for Microsoft customers​

Nothing changes in Copilot, Microsoft 365 or Azure today because of this filing. Enterprise buyers still have a reasonable question to ask. Many organizations deploying Copilot want to know how their vendor handles copyright risk, so cases like this feed procurement and legal reviews. It is worth checking what indemnification terms apply to your contract. This is general industry guidance, not something drawn from the complaint.

The case also shows how widely copyright litigation now reaches into the AI supply chain. Microsoft appears here as infrastructure provider, partner and Copilot distributor all at once. A court will ultimately decide how much responsibility each role carries.