.png)
AI Copyright Is Splitting Into Two Futures: Sue or License?
For years, the argument over artificial intelligence and copyright has sounded almost binary: either AI companies should be free to train on copyrighted material under fair use, or authors and publishers should be paid. This week, that argument became more interesting as we can now see two very different futures taking shape at the same time.
Quick Shortcuts
On 4 September, the Authors Guild and co-plaintiffs asked a New York federal court for summary judegment against OpenAI and Microsoft, arguing that copyrighted books were pirated, copied for training and transferred between the companies, and asking the court to reject the defendants’ fair-use defence. The Authors Guild’s filing puts the provenance of training material squarely back in the frame.
At almost the same moment, another creative industry offered a very different model. On 9 September, Reuters reported that AI music company Suno had launched new models in partnership with Warner Music Group and BMG, built around licensed participation and opt-in products for artists. The technology is different, but the principle matters to writers: creators do not necessarily have to choose between banning AI and giving their work away.
The real argument is shifting
The most important question for authors may no longer be simply whether AI can learn from books. It may be whether the companies building those systems have to account for how those books entered the training pipeline in the first place.
That distinction matters. A court could eventually decide that some forms of model training are transformative enough to qualify as fair use while still finding that obtaining material through piracy or unauthorised copying creates a separate legal problem. Those are not necessarily contradictory positions.
That tension is already visible beyond books. The Seattle Times and Newsday filed a new copyright suit on 4 September alleging that their journalism was copied without permission for AI training. Across publishing, journalism and music, the same issue keeps returning: where did the material come from, who had the right to provide it, and what obligations follow when that material becomes commercially valuable training data?
Why writers should care now
For working authors, this is not an abstract technology argument. It reaches directly into contracts, rights reversions, backlists, publishing agreements and the future value of a book after its first commercial life.
If licensing becomes a normal part of AI training, authors will need to understand which rights they have actually granted to publishers and which they have retained. A contract written ten or twenty years ago may never have contemplated machine learning. A book that has reverted to its author may sit in a very different legal position from one still controlled under an active publishing agreement.
That means the AI debate is moving out of the purely technical world and into the less glamorous but extremely important world of rights management. Who owns what? For how long? Who can license it? Who gets paid? And can a creator refuse?
There is a better middle ground
I have never found the simple pro-AI versus anti-AI division especially useful. It collapses too many different activities into one label.
I write my books. AI does not author them for me. But I do use AI around the creative business: research assistance, visualisation, marketing ideas, organisation and other supporting work. To me, those are fundamentally different things from another company ingesting the text of my novel into a commercial model without a clear agreement.
That distinction is why the licensing approach emerging in music is worth watching. It suggests that AI can remain a useful creative technology while recognising that the human work feeding commercial systems has value. It does not solve every problem, and books are not songs, but it moves the conversation beyond the false choice between unrestricted scraping and refusing the technology altogether.
What happens next could define the market
The US court battles will help establish the legal boundaries. But markets often move before courts finish speaking. If more AI companies decide that clean, licensed datasets are commercially safer and if creators see meaningful revenue from participating. Then then licensing could become a competitive advantage rather than merely a legal concession.
For authors, that would shift the practical question from “Will AI use my work?” to “On what terms?” That is a much healthier question because it gives writers something they have often lacked in this debate: agency.
There is also a transparency principle here that is hard to dismiss. Writers are increasingly being asked to disclose how AI has touched their work. Publishers, platforms and readers want provenance. It is reasonable to ask for the same standard in the other direction. If an AI company benefits from copyrighted books, authors should be able to understand how those books entered the system and under what authority.
If writers have to explain how AI touched our work, AI companies should have to explain how our work touched their models.
The question for authors
The next stage of AI copyright may not be decided by one sweeping ruling. We may end up with a patchwork: some training protected by fair use, some acquisition methods ruled unlawful, and a growing commercial licensing market operating alongside both.
That may be messier than either side wants. But it may also be closer to a workable future. One in which AI develops, authors retain meaningful rights, and legitimate access to creative work has a price.
So here is the question I think writers should be asking now: if AI training itself is eventually accepted as fair use, should companies still have to prove that the books they trained on were obtained lawfully and should authors have the right to license or refuse that use?
About Rob Frankson
Rob Frankson is a science-fiction author and creator of the Near Galaxy Saga. At 121 Minutes he writes about AI, authorship, publishing and the practical realities of being a working writer. More about Rob.
AI & Editorial Transparency
AI & The Author is edited and published by Rob Frankson. Artificial intelligence is used to assist with news research, initial drafting, content organisation and supporting imagery. All articles are reviewed and, where necessary, edited by Rob Frankson before publication. The opinions, editorial position and final decision to publish remain the author's.
Comments