Writers spent two years in federal court fighting to get paid for AI training data. Now some of those same authors are fighting a second battle, and this time the opponents are the agents and publishers who are supposed to be on their side. The dispute over who owns the right to settlement proceeds could hand 15 to 25 percent of any payout back to the publishing industry before authors see a dime.
The Settlement Fight Nobody Saw Coming
The AI copyright battles have been grinding through federal courts since 2023. Authors including Paul Tremblay and Mona Awad led early class action suits against major AI companies, arguing that training large language models on copyrighted books without licensing those works violates copyright law. According to court filings, Anthropic’s training data reportedly included millions of books sourced from shadow libraries.
According to the Authors Guild, more than 10,000 authors signed an open letter demanding that AI companies pay fair compensation for using their work. As the cases moved toward potential resolution in 2025 and 2026, a new conflict emerged inside the writing community itself. Literary agents and some major publishers began asserting contractual rights to a percentage of any settlement money authors receive.
Their argument is straightforward. Agency agreements typically grant agents 15 percent of an author’s earnings from their work. Publishers in some cases hold subsidiary rights that could apply to new revenue streams tied to those books. Authors disagree. Their position is that the settlement money compensates them for personal copyright infringement, not book sales, and that standard publishing contracts never contemplated AI training data as a revenue category.
Why This Is Really a Power Grab
I want to be direct about what this is. When a writer fights a two-year legal battle at personal risk, with no guarantee of winning, and then reaches a settlement, the people who sat on the sidelines shouldn’t automatically collect a percentage. That’s not how risk and reward are supposed to work.
The publishing industry has a long history of structuring contracts that extract maximum value from authors while offering minimum transparency. According to a 2024 Authors Guild survey, the median author income from writing in the United States fell below $20,000 for the first time in survey history. Publishers and agents are doing fine. According to Publishers Weekly industry estimates, the five major publishing conglomerates posted combined revenues of over $25 billion in 2024.
The agent commission dispute follows a pattern I see everywhere in business. The people who take no risk want to share in the reward. A rich mindset recognizes when a contract was written to serve someone else. A poor mindset assumes that whatever the contract says must be fair and stops there.
The authors pushing back are right. Their literary agents didn’t file briefs. Their publishers didn’t fund the litigation. The settlement money belongs to the people who fought for it.
This story matters well beyond publishing. Every creator who puts original work online is sitting on a copyright question that hasn’t been fully answered yet. If you make videos, write scripts, or produce content, your work has almost certainly been ingested somewhere without your consent. Tools like InVideo AI let creators produce video content at scale, and as the legal framework around AI training data firms up, those creators will want to understand exactly what rights they hold in their original output.
According to copyright attorney statements reported by the Nieman Lab, the central legal question is whether “copyright infringement damages” constitute “income from the work” under standard agency contract language. If courts say no, agents get nothing. If courts say yes, authors lose a significant chunk of what they won. That ruling could set a precedent that affects every future AI copyright settlement across every creative industry.
What This Means for You
If you’re a working author, here’s what I would do right now.
First, read every contract you’ve ever signed with an agent or publisher. Look specifically for language about “subsidiary rights,” “new media rights,” or “electronic rights.” Courts will look at that exact language to determine whether agents can claim AI settlement proceeds. If your contract is ambiguous, you want to know that before a settlement check is issued.
Second, don’t assume your agent is fighting for your interests in this dispute. Your agent has a financial stake in the outcome. That doesn’t make them your enemy, but it means their advice on this particular issue isn’t neutral. Consult an independent attorney, not one your agent recommends.
Third, if you’re a creator outside traditional book publishing, start documenting your original work now. Copyright registration costs under $65 for a group of unpublished works through the U.S. Copyright Office. That registration gives you standing to sue for statutory damages, which is the legal mechanism that gives these AI copyright cases their financial weight.
For independent creators and small publishers who want to build smarter content operations without giving up control to intermediaries, AppSumo regularly features lifetime deals on content and workflow tools that can help you own more of your stack and stay out of contracts that dilute your upside later.
The broader point is this: your intellectual property is an asset. Know what you own, know what you’ve licensed, and know who has a claim on future revenue from that work before someone else tells you.
The Bottom Line
Authors fought the AI companies and may actually win something. Now some of them have to fight their own representatives to keep it. This is the oldest story in business: you create the value, and then everyone with a contract tries to extract a percentage. The writers pushing back aren’t being difficult. They’re protecting what they earned. Every creator watching this fight should take notes, because the same question is coming for your work too.
Frequently Asked Questions
What is the Anthropic copyright settlement dispute about?
Authors sued Anthropic claiming the company used their copyrighted books to train its Claude AI models without permission or compensation. As those cases move toward resolution, a secondary dispute has emerged over whether literary agents and publishers can claim a percentage of any settlement money under existing contract terms.
Why are literary agents claiming part of the Anthropic settlement?
Standard literary agency contracts give agents a commission on earnings tied to an author’s work, typically 15 percent. Some agents argue that AI settlement proceeds fall under this commission structure. Authors counter that the money compensates for personal copyright infringement rather than book sales or traditional licensing income, so agency commissions shouldn’t apply.
Can publishers also claim a cut of an AI copyright settlement?
Possibly, depending on how the original publishing contract defined subsidiary or electronic rights. Courts will look at specific contract language to determine whether those provisions apply to a new category of revenue like AI training data compensation. Authors with tightly drafted, narrowly scoped contracts are in a stronger position to argue no.
What should authors do to protect their share of any settlement?
Authors should review their agent and publisher contracts now, looking for language about subsidiary rights and new media. They should also consult an independent attorney before any settlement discussions are finalized. Getting ahead of this question is far cheaper than litigating it after a check has already been issued.
Does the Anthropic settlement fight affect creators outside traditional publishing?
Yes. Any creator whose original work was used to train AI models has a potential copyright claim. Bloggers, screenwriters, journalists, and online content creators are all in a similar position. Registering your work with the U.S. Copyright Office gives you the legal standing to pursue statutory damages if you choose to act.
