Why the Anthropic Settlement Didn't Settle AI Copyright

The battle over what AI models can learn from copyrighted work is far from settled, and the legal landscape is more confusing than ever. Recent court decisions have sent mixed signals, leaving authors, tech companies, and lawyers wrestling with questions that copyright law, written for a pre-digital age, was never designed to answer.

A $1.5 Billion Settlement—With a Catch

Last year, Judge William Alsup ordered Anthropic to pay a $1.5 billion copyright settlement to a group of writers. On its face, the ruling appeared to be a win for authors, but the underlying decision actually favored Anthropic. Judge Alsup ruled that the AI company's training methods were lawful, and the penalty was for the company's use of pirated books from illegal online shadow libraries, not for training on copyrighted works.

In his ruling, Judge Alsup drew an analogy between AI training and human learning. As the judge wrote, “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.” This comparison, equating machine learning to a writer's study of literature, is a landmark in how courts might view AI training.

Attorney Cathy Gellis, an expert in intellectual property and technology law, sees the ruling as beneficial for AI companies. “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,” Gellis told TechCrunch. “Copyright law hinges on copying, but it doesn’t hinge on using the work or experiencing the work, consuming the work, reading the work.”

A Fine That Barely Registers

The $1.5 billion settlement might sound astronomical, but for a company projecting about $200 billion in annual revenue by 2028, it's a relative drop in the bucket. Gellis notes that the financial penalty is unlikely to deter AI developers, especially when the underlying legal precedent is in their favor.

The sheer scale of the AI industry means that even massive fines can be absorbed as a cost of doing business. This raises the question: what kind of legal deterrent actually works when a company's future revenue dwarfs any potential penalty?

Fair Use: The Heart of the Matter

These legal battles hinge largely on the doctrine of fair use, which allows for using copyrighted material without permission under certain circumstances. Courts evaluate factors like the purpose and nature of the use, the amount used, and the impact on the market for the original work.

Jason Henderson, Senior Attorney and Founder of the IP & Media Practice at JWL International, explains that the courts are “all over the place” in their reasoning. “Everybody is very worried right now because the law is all over the place, and it’s because of this question,” Henderson told TechCrunch. “They know that the AI model has been trained on so much stuff, and the law has not really caught up to that question.”

“Copyright is always about protecting and growing the market,” Henderson added. “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.”

When Training Is Not Transformative

The trend Henderson describes is visible in the case of Thomson Reuters v. Ross Intelligence. Thomson Reuters sued Ross Intelligence for copying its legal content to build a competing AI-based legal platform. In that case, Judge Stephanos Bibas ruled against Ross, writing, “Ross’s use is not transformative because it does not have a ‘further purpose or different character’ than Thomson Reuters’s.”

The court found that Ross's use was not fair use because it directly competed with the original content. This contrasts with the Anthropic case, where the judge viewed the training as more analogous to reading and learning, even if the source material was pirated.

For authors, the potential argument that AI chatbots could replace them by generating synthetic books has not yet found success in court. The outcomes so far suggest that the purpose of the AI's use—whether it directly competes with the original work—is a crucial factor.

Copyright Law: Stuck in 1976

Adding to the complexity, copyright law itself is outdated. The Copyright Act of 1976 has not been updated since its passage, meaning judges are interpreting guidelines from 50 years ago in the context of modern AI. This creates a patchwork of decisions as courts try to apply outdated rules to novel questions.

“The law has not really caught up to that question,” Henderson said, emphasizing the uncertainty facing AI companies and creators alike. Without clear legislative guidance, every new case sets precedent, but those precedents can conflict across different jurisdictions.

AI-Generated Works: A New Can of Worms

The issue extends beyond training data to the question of whether AI-generated content itself can be copyrighted. In Thaler v. Perlmutter, the court ruled that a work that is 100% AI-generated is not copyrightable. This opens a host of new questions about how to prove whether a work was created by AI and what percentage of human involvement is required.

“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,” Gellis said. “[AI] is forcing us to look at a whole bunch of decisions that we kind of ignored for a while.”

The line between human and machine authorship is blurring, and the law is struggling to keep up. Who owns the rights to a poem written by an AI trained on the works of living poets? What if a human edits the AI's output?

Pending Litigation and Uncertain Future

Most AI companies are still embroiled in ongoing litigation, meaning there is no definitive answer on the horizon. Each new ruling shapes the legal landscape, but conflicting decisions across courts could lead to a prolonged period of uncertainty.

“What you are seeing is that the initial opening volleys are being influential, and that influence itself could be undone if other courts decide different things, and it’ll take later states of litigation to figure out which one will prevail,” Gellis said. “But in the meantime, all these decisions are shaping everything that’s happening. It would be kind of foolish for the AI companies to ignore them.”

Gellis also highlighted the emotional stakes: “I think one of the issues with this entire area of law and this entire area of technology is there’s a lot going on. It’s very complex and there are a lot of raw feelings about what is happening, both for and against.”

Why This Matters for Everyone

The outcome of these legal battles will affect not just AI companies and authors, but anyone who uses or consumes AI-generated content. If AI training is deemed fair use, it could accelerate the development of powerful tools that might disrupt creative industries. If it's not, it could slow innovation and leave the legal status of countless AI models in question.

For authors and publishers, the stakes are existential: their livelihoods depend on the ability to control and profit from their work. For AI companies, the cost of litigation and potential damages could reshape their business models. And for the public, the answers to these questions will determine what kinds of AI tools are available and how they are built.

One thing is clear: the law is lagging behind technology, and until it catches up, we're in for a messy, case-by-case free-for-all. As Henderson put it, “They know that the AI model has been trained on so much stuff, and the law has not really caught up to that question.”