The complaint filed Friday carries the tone of an elegy with a legal caption. The Seattle Times and Newsday, the Long Island daily, accuse OpenAI and Microsoft of scraping their articles without permission to train ChatGPT and Copilot, and they describe the technology in terms borrowed from myth: a snake eating its own tail, consuming the work of the institutions whose reporting supplied the material for its intelligence. Left to continue, the papers argue, the practice would destroy the organizations that produce journalism, and the damage would arrive too quickly to repair.
The suit joins a legal campaign that has been building since late 2023. The New York Times was first, suing OpenAI and Microsoft in December of that year, and it has since been followed by other newspaper groups, digital publishers, authors and artists, with many of the cases consolidated before a federal judge in New York. Friday’s filing adds two regional papers with long histories. The Seattle Times is one of the last family-owned metropolitan dailies in the country, and Newsday has covered New York’s suburbs for decades; both have spent years moving readers to digital subscriptions while watching traditional revenue shrink. They are not asking the court to stop AI. They are asking to be paid for the role their work played in building it, and they frame the case around a simple claim: no permission was sought, and no license was offered.
The case has an awkward feature the defendants may find hard to explain away. OpenAI and Microsoft have both supported journalism projects at The Seattle Times, and the paper’s leadership has publicly urged the industry to experiment with AI. The relationship did not prevent the suit. A Microsoft spokesperson said the company was surprised by the filing and remains willing to talk, a response that hints at the settlement math already visible in earlier cases.
The legal questions are familiar by now. OpenAI and Microsoft are expected to argue that training on published text is a fair use, a transformation of the material rather than a reproduction of it, and that their systems quote little of what they learn. The newspapers will argue that ingestion of millions of articles is copying on an industrial scale, that the models compete with the publications that produced their training data, and that no amount of transformation excuses taking work without payment. Courts have allowed cases of this kind to proceed, but no appellate ruling has settled the central question, and the answer may take years to arrive.
The newspapers’ complaint also targets the way the models reach readers. ChatGPT answers questions with summaries that can stand in for an article, and Microsoft sells Copilot inside the browsers and office software where people once encountered journalism, the papers argue. Each time a user gets an answer assembled from reporting without visiting the site that paid for the work, the economics of a newsroom weaken by a little; the complaint says the defendants built that behavior into their products deliberately rather than stumbled into it.
Microsoft occupies an unusual position in the dispute. The company has invested billions in OpenAI, hosts its models on its cloud and sells its own AI products, and when papers sue the pair together they force a question the industry would rather keep separate: how much responsibility belongs to the investor and distributor that put the technology into the world’s offices. Both companies have promised customers they will cover copyright claims arising from their tools, promises that make the underlying litigation a financial question for the defendants themselves, not just for the papers.
The larger context is a widening split inside the news business. Some of the biggest publishers in the world have signed licensing deals with OpenAI and other AI companies, trading access to their archives for payments and product partnerships. Others, including most of the American newspaper industry, have chosen litigation, arguing that deals reward the takers and undersell the value of journalism. The split has produced strange alliances and pointed criticism in both directions. For smaller outlets, the choice has often been different, analysts note: no deal was ever offered, and their archives are already inside the models.
Regulators are watching the fight with more attention than they gave the early rounds. Congressional hearings on AI and copyright have pressed companies to account for their training data, and the White House has weighed rules that would require disclosure of what models are built on. The newspapers’ suit gives lawmakers a concrete pair of defendants and a concrete set of claims to point at, whatever the courts eventually decide.
For the two papers, the suit is partly about money, but the language suggests it is about something else as well: putting a record on file. A complaint that calls the defendants’ technology a self-devouring snake is written for judges, for the public and for the history of the argument. Whatever the outcome, the papers have stated on the record that their work was taken without permission, and that they believe the taking, left unanswered, would end them. In an industry that has watched its revenues fall for two decades, that is not hyperbole; it is a description of the fear the business lives with.
