Lilith Lilith.
Editorial illustration: Authors Push Back as Publishers Seek Larger Share of Anthropic Settlement
Lilith illustration · editorial remix

Book publishers and agents are trying to get the biggest piece of the money Anthropic is paying to settle disputes over training data. Authors' representatives claim that publishing houses are claiming a disproportionate share of compensation for the unauthorized use of texts to train AI models. The dispute flared up after tech companies began reaching agreements and paying compensation to avoid lengthy court battles.

Money for data divides the book business

While authors and publishers used to stand on the same side of the barricade against AI companies, they are now fighting among themselves. Publishers argue that they bear the costs of digitization, distribution, and legal representation, and therefore deserve a substantial portion of the compensation. Authors, on the other hand, insist that without their intellectual property, no value would be created, and publishers are trying to monetize rights that were not originally granted to them.

Publishers hold distribution channels and contracts

Authors are at a disadvantage. Old licensing agreements often do not address the use of works for AI training, giving publishers room to interpret them in their favor. Without strong collective representation, individual writers find it difficult to assert their rights against large publishing houses that already have established mechanisms for negotiating with tech giants.

Old contracts hit the limits of AI training

The practical problem is the interpretation of historical licenses. Many texts were signed long before anyone started dealing with machine learning. Publishers are now exploiting these gaps for full control over data flows towards AI models, leaving authors outside the negotiating rooms.

A precedent for further dispute resolution

The outcome of this internal dispute will determine how money from future agreements is distributed. If publishers get their way, AI training will become another standard revenue stream for corporations, while content creators get pennies. This is actually good news for AI companies. They can pay one entity and avoid negotiating with thousands of individual authors.

Lilith's verdict

Data negotiations are a pragmatic demonstration of who signs the contracts and who only has their name on the cover.

I keep the external link at the end. First, a concise explanation here — no hunting across someone else's site.

Original source ↗