In June 2025, a US federal judge used the word "spectacularly" — not to describe the harm done to authors whose books were fed into an AI model, but to describe how transformative that use was, and therefore how legal. Judge Alsup's ruling in Bartz v. Anthropic is one of only two federal decisions that directly address whether AI companies can train on copyrighted books without permission, and if you've been getting your legal updates from Twitter threads and outraged quote-tweets, there's a strong chance you've absorbed a version of this story that doesn't match the court record. That gap between perception and reality is now one of the biggest strategic risks facing indie authors.
Here's what actually happened. In Bartz v. Anthropic, the court issued a split decision. Training on copyrighted books that Anthropic had legally acquired? Fair use. The judge even said digitizing print copies for that purpose was permissible as a format change. But — and this is the part the rage-posts leave out — the court also found that retaining a library of pirated copies constituted copyright infringement. A second case involving Meta reached a similar pro-training conclusion but on different legal grounds. These are trial-level rulings, not Supreme Court precedent, and other courts could go the opposite direction. As Jane Friedman has meticulously documented in her ongoing AI legal FAQ, the early rulings are not final, and several critical questions remain wide open: How will courts measure market harm from AI outputs that compete with human-authored works? What rights do authors have when companies buy their books specifically for training with no license? There's no settled answer yet. But the direction of travel in these first two rulings should sober anyone who assumed the law would automatically side with creators.
My position is blunt: indie authors who are waiting for the courts to save them are making a dangerous bet. The legal system is moving slowly, the rulings so far favor AI companies on the training question, and even when authors win on specific points — like the piracy finding in Bartz — the practical payoff is uncertain. The ALLi news team reported in April 2026 that in the Anthropic settlement, reduced legal fees may actually increase payouts for authors, which sounds encouraging until you realize it's an acknowledgment that the legal fees were threatening to swallow the recovery entirely. Meanwhile, a Supreme Court ruling in Sony vs. Cox is being watched for its potential influence on AI copyright cases, adding yet another layer of unpredictability. The legal pipeline is long, expensive, and offers no guarantees.
So what should you actually do with this information? First, stop treating social media as a legal research tool. Friedman's observation is worth repeating: if you primarily learn about AI and copyright from social media, you are likely misinformed. Even literary agents and editors get the basics wrong. The misinformation isn't just annoying — it leads to bad business decisions, like assuming your copyright automatically prevents any AI use of your work, or believing that disclosure requirements and AI detection tools are more robust than they are. Second, recognize that the unresolved questions — how much human involvement is "sufficient" to copyright AI-assisted content, how publishing contracts will evolve to address AI rights — are not abstract policy debates. They will directly shape your income, your contracts, and the value of your backlist within the next few years.
The real competitive advantage for indie authors right now isn't legal protection — it's strategic awareness. Understanding the difference between what courts have actually decided and what you wish they'd decided lets you make smarter choices about contracts, licensing, platform terms, and how you build reader relationships that no AI model can replicate. The authors who thrive through this transition won't be the ones who screamed loudest on social media. They'll be the ones who read the rulings.
Here's your one action item: go to Jane Friedman's AI copyright FAQ — she keeps it updated as new rulings drop — and read the actual legal distinctions she outlines, particularly around fair use, AI training, and what qualifies for copyright registration. Bookmark it. Check it quarterly. Then review every publishing contract you sign this year for AI-related clauses, because publishers are already inserting them, and the language matters more than most authors realize. The law won't protect you by default. Your informed decisions might.