Two lawsuits against AI companies made news this week, but the more interesting legal fight is happening behind the scenes: who actually gets paid when publishers settle copyright claims over AI training?

The Seattle Times and Newsday filed suit against OpenAI and Microsoft on Friday, joining the growing list of publishers suing over AI training. The complaints follow the familiar pattern: we own copyrights, you trained on our content without permission, pay us. Standard stuff at this point.

But over at the Anthropic settlement, something messier is unfolding. Authors are pushing back against publishers and literary agents who are making claims on settlement payments that were supposed to go to writers.

The Problem With Publishing Contracts

Here’s what’s happening. When publishers and authors’ groups reached a settlement with Anthropic earlier this year, the money was meant to compensate authors whose work was used to train Claude. But now publishers are arguing that their contracts give them rights to a portion of those payments.

This isn’t a new trick. Standard publishing contracts give publishers certain rights to exploit the work. The question is whether “exploit” includes getting paid when an AI company settles a copyright lawsuit over training data.

Publishers think it does. They’re pointing to broad language in contracts about subsidiary rights and new technologies. Authors think that’s absurd. The settlement is compensation for infringement, not a book sale or licensing deal.

Who Has the Better Argument?

The authors probably do, but it depends on the specific contract language.

Most publishing agreements were written before anyone imagined AI training as a commercial use. The contracts talk about adaptations, translations, film rights, that kind of thing. Trying to shoehorn AI training settlements into those categories is a stretch.

The stronger argument for authors is that settlement payments are compensating for harm to their copyrights, not exploiting the work in the way a contract contemplates. If someone steals your car and later pays damages, your insurance company doesn’t get a cut just because they covered collision.

But publishing contracts are notoriously publisher-friendly, and some contain language broad enough that a court might side with the publishers anyway. The real answer is going to come down to how specific contracts define subsidiary rights and whether settlement payments count as licensing revenue.

What This Means for Future Settlements

The fight matters because it will shape every AI copyright settlement going forward. If publishers can claim a share of author settlements, writers will see less money. That might also discourage settlements, since authors won’t want to agree to deals that mostly benefit their publishers.

It’s also going to affect how future publishing contracts get written. Expect authors’ advocates to push for explicit language clarifying that settlement payments for infringement belong to the author, not the publisher.

The larger question is whether the existing settlement structure even makes sense. Most of these deals are being negotiated by publishers and authors’ groups, not individual writers. That made administrative sense when Anthropic wanted to settle quickly, but it’s created this exact problem: who speaks for authors, and who gets paid?

The Lawsuits Keep Coming

Meanwhile, the Seattle Times and Newsday suits against OpenAI and Microsoft won’t be the last. The legal theory is well-established at this point: publishers own copyrights in their articles, AI companies trained on those articles without permission, that’s infringement.

OpenAI will argue fair use, as it has in every other case. The company will say that training is transformative, doesn’t harm the market for the original works, and serves the public interest. Courts haven’t ruled definitively on that argument yet, which is why we keep seeing settlements.

What’s notable about these new suits is that they’re coming from regional and mid-sized outlets, not just major publishers. That suggests the legal strategy is spreading beyond the biggest players who can afford extended litigation.

It also means more opportunities for settlement fights like the one playing out over Anthropic. If the Seattle Times or Newsday settle, their reporters will want to know whether the paper gets to pocket the money or whether individual writers get a share. And if the answer is that the publisher keeps everything, expect more pushback.

The AI copyright wars have been framed as a fight between publishers and AI companies. Turns out there’s another battle happening inside the publishing industry itself.