The Seattle Times and Newsday sued OpenAI and Microsoft over training data

The Seattle Times is an unusual plaintiff, because Microsoft and OpenAI have funded some of its journalism projects and fellowships. A regional paper is suing the companies that helped pay for parts of its newsroom.

That is less contradictory than it appears, and it is the most honest thing about the case. Grant funding for a fellowship is philanthropy, revocable and small. The claim is about a structural transfer: reporting produced at cost by a newsroom, ingested at no cost, and used to build a product that answers the reader's question without sending them to the newsroom.

One of those is a donation. The other is the business model. A publisher can accept the first and still litigate the second, though it makes for uncomfortable filings.

The regional paper's position is weaker

The New York Times, which filed against the same defendants in December 2023, has leverage a regional paper does not: a global brand, a subscription business that works, and archives valuable enough that licensing them is a real alternative to suing over them.

The Seattle Times and Newsday have none of that. Regional news economics were broken by search and social before AI arrived, and their realistic outcomes are a modest settlement, a licence at a price set by someone else's litigation, or nothing.

Which is why these suits keep being filed. For a regional publisher, joining the queue costs contingency-fee legal time and preserves a claim on whatever the Times case establishes. Staying out forfeits it.

What the government just did to the odds

Three days before this filing, the Justice Department submitted a statement supporting OpenAI and Microsoft's position that training on copyrighted text is fair use, timed ahead of summary judgment motions in the Times case.

That does not bind the court, and it does not address every theory in the newspaper complaints. But it means these plaintiffs are now arguing against the defendants and the executive branch's reading of the statute at the same time.

The distinction that will decide it

The complaints blur two different acts, and the outcome probably depends on separating them.

Training is the historical act: ingesting an archive to learn statistical structure. The transformative-use argument is strongest here, and it is what the DOJ filing supports.

Retrieval is the live act: a product fetching today's article, summarising it and answering the reader's question. That directly substitutes for the visit that funds the reporting, and it is a much harder fair use case.

The newspapers have the better argument on the second and the weaker one on the first, and their complaints lead with the first.