Idealism meets pragmatism as the Swedish Writers’ Union gathers its members for a meeting about the AI future. The mood is set to be tense — but the central conflict has already been partly decided.

It’s not surprising that the tech world is so disliked across large parts of society. Every new technology shift comes with promises of riches and a brighter future — last time it was social media, this time it’s AI.

The riches, however, mostly seem to stay in Silicon Valley. And the broader benefits leave plenty of question marks too. It’s a long way from Google’s Nobel-winning Alphafold project — the AI system that can compute protein structures — to using enormous amounts of electricity to generate strange memes and AI slop.

It’s understandable, then, that Swedish authors aren’t cheering when the conversation turns to their works being included in a new Swedish AI model. The goal is to produce a language model that understands Swedish deeply and that isn’t dependent on American tech companies.

The technology has underdelivered before. When the Swedish Writers’ Union gathers on Saturday to discuss the issue, the mood is set to be sour.

At the same time, there seems to be a misunderstanding both about how the technology and the market function, how they could function — and what’s actually happening right now. Whether you like it or not.

One of the project’s critics, the poet Olivia Bergdahl, tells Aftonbladet: “The way these machines work, it’s the books that get used, but really it becomes more a question of what price tag you put on the language an author has developed — and that’s impossible.”

That may be true. But the price tag on the existing works that have found their way into today’s AI models is already set. It’s exactly zero kronor. So we’re not talking about a situation where authors’ books might end up in AI models — they’re already there. Along with a long list of other copyrighted material that has also ended up there, without permission.

Last autumn, AI company Anthropic reached a settlement in which it agreed to pay 1.5 billion dollars for having used books without permission to train its AI models. The examples are many, and they exist in adjacent industries too. The New York Times, for instance, has sued OpenAI for taking its journalism in a case still working its way through the American legal system.

The conversations the Swedish Writers’ Union is trying to have about a new Swedish language model can be seen in this light. The initiative looks less like a way to embrace AI technology and more like an attempt to step a couple of paces back from where we are today. The aim is to be able to regulate compensation and the use of books in these models under far more controlled terms than what’s happening now.

If this sounds like a kind of hostage situation, there’s a reason for that. American and Chinese companies are currently taking Swedish and foreign literature and training their AI models on it. No one — neither the authors themselves nor the Writers’ Union — has any say in that fact. That obviously doesn’t mean you have to welcome this future with open arms and simply accept it. But to realistically argue for staying entirely outside it, you have to think about how that would actually work in practice. It might mean suing the AI companies in question. But even that is something you’d most likely want to do as a collective rather than as an individual author.

The Swedish Writers’ Union has, thanklessly, ended up taking the blame for this messy AI era. What they’re discussing is participating in a Swedish research project where a different kind of control around compensation and copyright could be built.

If members decide — now or in the future — that they want to fully withdraw from all AI involvement, there’s the option of finding solutions for that too. And if large-scale copyright infringement lawsuits also arrive in Sweden, it might be wise to have a partner who understands how the technology works from the inside.

That holds — unfortunately, you might say — even for those who want nothing to do with it.

This analysis was first published in SvD Näringsliv, in Swedish, on May 29th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.