The Seattle Times and Newsday ask a court to destroy the models trained on their journalism

The Seattle Times Company and Newsday LLC sued OpenAI and Microsoft on September 4, 2026, in the U.S. District Court for the Southern District of New York. The complaint, docketed as 1:26-cv-07644, names nine OpenAI entities alongside Microsoft Corporation, runs to 38 pages, and demands a jury trial. It joins a line of newspaper suits that the same court has already gathered into one consolidated proceeding.

What the complaint charges

The seven counts fall into three groups. Two are copyright counts under section 501 of the Copyright Act, one for direct infringement and one for vicarious infringement. Two more are Digital Millennium Copyright Act counts under section 1202(b). They concern what the papers describe as the removal of copyright management information: article titles, author names, and copyright notices. The complaint says that removal happened both when their work was collected and when it comes back out of the products.

The last three counts are about trademark dilution, one federal and two under state law. They rest on a different theory from the copyright counts. The papers argue that when the products invent content and attribute it to The Seattle Times or to Newsday, the damage lands on the name of the paper rather than on any single article.

• Copyright infringement and vicarious copyright infringement, section 501.
• Removal of copyright management information, and distribution of works with it removed, section 1202(b)(1) and section 1202(b)(3).
• Federal trademark dilution, section 1125(c) of the Lanham Act.
• State dilution claims under Washington's trademark dilution statute and under New York General Business Law section 360-L.

The evidence the papers put on the record

The complaint prints model output beside the original text. In one example it says a defendants' model returned 88 consecutive words of a Seattle Times article verbatim, from a prompt that contained only the headline, the publication date, and the address of the piece. The papers draw the inference themselves: the words came out of training, not out of anything a user had pasted in.

The article in that example belongs to the Pulitzer-winning series on the Boeing 737 MAX crashes. The complaint also points at OpenAI's own published description of the WebText dataset, where seattletimes.com sits at number 221 among the thousand domains listed.

Both papers had closed the door well before any of this. Newsday's robots.txt tells OpenAI's crawler and Common Crawl to stay out. Its terms of service, in force since October 12, 2023, prohibit using its content to train a machine learning or artificial intelligence system. Newsday has shown non-subscribers no editorial text at all since August 2022. The Seattle Times has charged for access since the spring of 2013, and at the time of filing a non-subscriber got a single free pageview.

The remedy is the part worth watching

The money is the ordinary part. Under the copyright counts the papers elect between statutory damages and actual damages plus the defendants' profits, and under the dilution counts they ask for an accounting of profits, treble damages, and attorneys' fees.

The unusual request sits in paragraph (d) of the prayer for relief. It asks the court to order impoundment or destruction, under section 503 of the Copyright Act, of every copy of the plaintiffs' works. The same request covers all large language models and training datasets that incorporate those works or derivatives of them.

This court already handles cases of the same shape. The complaint notes that OpenAI and Microsoft accepted its jurisdiction in the suits brought by The New York Times, Raw Story, The Intercept, and the Daily News. Those cases are consolidated as In re OpenAI, Inc. Copyright Infringement Litigation, No. 25-md-03143.

What has not happened

No court has ruled on any of this. A complaint is one side's account, filed before the other side has answered, and the fair use question underneath every case of this kind is still open in this district. The Verge reported that neither OpenAI nor Microsoft responded immediately to a request for comment. TechCrunch reported that Microsoft said it was surprised by the suit and would be happy to discuss a solution.

Four exhibits came in with the complaint: the lists of registered copyrighted works for each paper, and the trademark registration certificates they rely on. Everything else in the case is still ahead of it.