Last year's $1.5 billion copyright settlement against Anthropic seemed like a moral victory for writers, but the judge's actual ruling on AI training legality may have handed a significant advantage to the AI industry. Judge William Alsup determined that training AI models on copyrighted books is lawful under copyright law, penalizing Anthropic only for pirating those books from illegal shadow libraries rather than for the training itself.

What Did Judge Alsup Actually Rule on AI Training?

The distinction matters enormously for the future of AI development. Alsup compared how large language models (LLMs), the neural networks that power ChatGPT, Gemini, and Claude, ingest trillions of words to how human writers study literature to improve their craft. "Like any reader aspiring to be a writer, Anthropic's LLMs trained upon works not to race ahead and replicate or supplant them, but to turn a hard corner and create something different," Alsup wrote.

This framing is crucial because copyright law fundamentally hinges on copying, not on consuming or reading a work. The judge's reasoning suggests that training an AI model on copyrighted material is more akin to reading and learning from a book than to copying it, which is a legally protected activity. For context, Anthropic is projecting roughly $200 billion in annual revenue by 2028, making the $1.5 billion fine a manageable business expense rather than an existential threat.

"I think it is generally good news for AI training that he looked at what was going on and really sort of thought it analogous to reading a copyrighted work as opposed to copying a copyrighted work," said Cathy Gellis, an attorney with expertise in intellectual property, copyright, and technology.

Cathy Gellis, Attorney specializing in intellectual property, copyright, and technology

Gellis emphasized that copyright law doesn't hinge on using, experiencing, consuming, or reading a work. It hinges on copying. This distinction could reshape how courts evaluate future AI training cases, potentially making it harder for authors to argue that AI companies are infringing their copyrights simply by using their works as training data.

Why Is Copyright Law Struggling to Keep Up With AI?

The core problem is that copyright law hasn't been meaningfully updated since 1976. Judges are forced to interpret guidelines written 50 years ago when confronting legal questions that could shape the entire future of the AI industry. This legal vacuum has created widespread uncertainty about what is and isn't permissible when training AI models.

Most of these disputes hinge on "fair use," a carve-out in copyright law that allows people to use copyrighted materials without explicit permission for purposes like criticism, parody, education, and commentary. Courts evaluate fair use by considering several factors:

• Purpose and Nature: Whether the use is transformative, meaning it adds new meaning, message, or expression to the original work.
• Amount Used: How much of the original copyrighted work was incorporated into the new creation.
• Market Impact: Whether the new use harms the original work's market value or the copyright holder's ability to profit from it.

"Copyright is always about protecting and growing the market. The courts are kind of all over the place in their reasoning in AI cases. What's tending to win is if what you're doing is you're training on somebody's property because your purpose is to directly compete, then the courts will frown on it. If what you're doing is not going to compete, then the courts are tending to find ways that it will be okay," said Jason Henderson, Senior Attorney and Founder of the IP & Media Practice at JWL International.

Jason Henderson, Senior Attorney and Founder of the IP & Media Practice at JWL International

When Do Courts Say AI Training Crosses the Line?

Not all AI training cases result in fair use rulings. In a case involving Thomson Reuters and Ross Intelligence, the courts drew a clear line. Ross Intelligence had copied Thomson Reuters' content to build a competing AI-powered legal research platform. Judge Stephanos Bibas ruled that this was not fair use because Ross's use did not have "a further purpose or different character" than Thomson Reuters's original work. The key difference: Ross was directly competing with Thomson Reuters using its own training data.

This precedent suggests that courts may be more sympathetic to AI companies when they're not directly competing with the copyright holders whose works they used for training. However, authors could potentially argue that chatbots powered by their works are competing with them by generating synthetic books and content. So far, that argument has not prevailed in court.

How to Navigate the Current AI Copyright Uncertainty

• Monitor Pending Litigation: Most AI companies are currently lodged in pending copyright litigation, meaning definitive legal standards won't emerge for months or years. Early court rulings are shaping industry behavior, but later cases could overturn them.
• Understand the Distinction Between Training and Output: Copyright questions about training AI models are entirely separate from questions about copyrighting AI-generated content. The Thaler v. Perlmutter case ruled that 100% AI-generated works are not copyrightable, creating new questions about how to prove whether a work was AI-generated and to what degree.
• Recognize the Broader Implications: As attorney Gellis noted, AI is forcing courts and lawmakers to reconsider decisions they've ignored for decades. For example, when you write a novel in Microsoft Word and use spell check, we don't assume Word owns your novel. AI is raising similar questions about the degree of tool assistance that still preserves human authorship.

"If you write your novel in Microsoft Word and run spell check, we kind of feel comfortable with the idea of saying that Word does not own your novel. AI is forcing us to look at a whole bunch of decisions that we kind of ignored for a while," said Cathy Gellis.

Cathy Gellis, Attorney specializing in intellectual property, copyright, and technology

The reality is that the legal landscape for AI copyright will remain in flux for years. Gellis explained that the initial court rulings are influential, but that influence could be undone if other courts decide differently. It will take later stages of litigation to determine which legal reasoning ultimately prevails. In the meantime, all these decisions are shaping everything happening in the AI industry.

For AI companies, ignoring these early rulings would be foolish. For authors and publishers, the current moment represents a critical window to shape how courts interpret copyright law in the age of AI. The outcome will determine whether training on copyrighted works remains a foundational practice for building powerful AI models or becomes legally and financially prohibitive.