AI, Copyright, and the Evolving Legal Battle Over Creative Works

The rapid advancement of artificial intelligence, particularly large language models (LLMs) like ChatGPT, Gemini, and Claude, has ignited a complex and contentious debate surrounding intellectual property rights. These powerful AI models are trained on colossal datasets, often described as "seemingly infinite," encompassing hundreds of millions of books, online articles, academic papers, and virtually all accessible content on the internet. This extensive training process raises profound questions, as a vast number of published authors have, without their explicit knowledge or consent, inadvertently contributed their creative works to the development of the very AI tools that now pose a significant threat to their livelihoods. To many, this situation appears to be a clear violation of existing legal frameworks, specifically copyright law. However, the legal landscape governing AI training and content generation is far from straightforward.

Cathy Gellis, an attorney specializing in intellectual property, copyright, and technology, highlighted the inherent complexity of this emerging legal area. "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," Gellis told TechCrunch. "It’s very complex and there are a lot of raw feelings about what is happening, both for and against." This sentiment underscores the profound impact AI is having on established legal norms and the deeply personal implications for creators.

One of the most significant early rulings addressing this issue involved the AI company Anthropic. Last year, in a decision that garnered considerable attention, Judge William Alsup ordered Anthropic to pay a substantial $1.5 billion copyright settlement to a consortium of writers. These authors alleged that their works had been used without permission to train Anthropic’s AI models. While the sheer size of the settlement initially appeared to be a resounding moral victory for authors, signaling judicial protection for their creative endeavors, a closer examination of Judge Alsup’s ruling reveals a critical nuance. The judge explicitly ruled that Anthropic’s AI training itself was lawful. The hefty penalty was not levied for the act of training on copyrighted material, but rather for the method by which Anthropic acquired those materials – specifically, pirating books from illegal online "shadow libraries." This distinction is pivotal for understanding the current legal trajectory.

Judge Alsup drew a compelling analogy to explain his reasoning regarding the lawfulness of AI training. He 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." In this comparison, the judge likened an LLM’s ingestion of trillions of words to a human writer’s extensive study of literature. The implication is that the AI’s process of learning from and synthesizing information, rather than direct replication, falls outside the traditional definition of copyright infringement.

Gellis views this ruling as significantly more advantageous for AI companies than for authors. She questioned the true impact of a $1.5 billion fine on a company like Anthropic, which is projected to achieve an astounding $200 billion in annual revenue by 2028. From her perspective, the financial penalty, while substantial, is a relatively minor setback compared to the legal precedent established regarding the training process itself. "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 explained. She emphasized a fundamental principle of copyright law: "Copyright law hinges on copying, but it doesn’t hinge on using the work or experiencing the work, consuming the work, reading the work." This interpretation suggests that the act of an AI model processing copyrighted data for learning purposes, without creating direct copies for distribution, might not constitute infringement under current law.

A major contributing factor to the legal ambiguity is the age of the governing legislation. The Copyright Act of 1976, which forms the bedrock of intellectual property law in the United States, has not been substantially updated since its enactment. This means that judges today are tasked with interpreting guidelines from nearly half a century ago – a period predating the internet, digital content, and certainly advanced artificial intelligence – to resolve complex legal questions that could profoundly shape the future of the entire AI industry.

Jason Henderson, Senior Attorney and Founder of the IP & Media Practice at JWL International, articulated the widespread concern within the legal community. "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." This disconnect between technological reality and legal frameworks creates significant uncertainty for both creators and AI developers.

Many of these legal questions hinge on the doctrine of "fair use," a crucial exception within copyright law. Fair use determines whether the use of a copyrighted work without explicit permission is legally permissible. At its core, fair use is designed to protect and encourage activities such as criticism, commentary, news reporting, teaching, scholarship, and research – essentially, the ability to build upon and iterate on existing creative works for a "further purpose or different character." When evaluating a fair use claim, judges typically consider four key factors:

1. The purpose and character of the use: This includes whether the use is commercial or for non-profit educational purposes, and critically, whether the new work is "transformative." A transformative use adds new meaning, expression, or message to the original work, rather than merely superseding it.
2. The nature of the copyrighted work: This factor considers whether the original work is factual or creative, published or unpublished.
3. The amount and substantiality of the portion used: Judges assess how much of the copyrighted work was used and whether the portion taken was central to the original.
4. The effect of the use upon the potential market for or value of the copyrighted work: This factor examines whether the new use harms the market for the original work or for derivative works based on it.

Henderson emphasized that "Copyright is always about protecting and growing the market." He noted the inconsistency in judicial reasoning in AI cases, but identified a prevailing trend: "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."

This distinction was clearly illustrated in the case where the media and technology giant Thomson Reuters sued the legal research firm Ross Intelligence. Ross Intelligence had copied content from Thomson Reuters to develop its own competing, AI-based legal platform. Judge Stephanos Bibas ruled against Ross Intelligence, stating, "Ross’s use is not transformative because it does not have a ‘further purpose or different character’ than Thomson Reuters’s." In this instance, the court found that Ross Intelligence’s actions constituted a direct market substitute, thus failing the fair use test. While authors could logically argue that generative AI chatbots, by creating synthetic books and articles, are directly competing with human creators, this specific argument has not yet definitively prevailed in court in the same way the Thomson Reuters case did. The distinction lies in whether the AI training is seen as creating a new tool for a different purpose (like Anthropic’s general LLM) or directly replicating and competing with the original work in its market (like Ross Intelligence).

Cathy Gellis points out that the relationship between AI and copyright needs to be viewed through two distinct lenses: the copyright implications of AI training and the copyrightability of AI-generated content. The legal considerations for each are quite different.

Regarding the copyrightability of AI-generated content, a landmark case, Thaler v. Perlmutter, delivered a significant ruling. The court determined that if a work is 100% AI-generated, it is not copyrightable. This decision is rooted in the long-standing principle of "human authorship" in copyright law, which traditionally requires a human creator. This ruling, however, immediately opens a "whole new can of worms." It raises complex practical questions: How can legal systems definitively prove whether a work was generated using AI? And if AI was involved, how can the percentage of AI contribution versus human input be accurately determined?

Gellis used a relatable analogy to explain this challenge: "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." This highlights the intuitive understanding that tools assist human creativity without claiming ownership. However, the sophistication of AI tools blurs this line dramatically. Gellis noted, "[AI] is forcing us to look at a whole bunch of decisions that we kind of ignored for a while." The legal system must now grapple with defining the threshold of human involvement necessary for copyright protection in an era where AI can produce highly complex and original-seeming works.

Currently, most major AI companies remain entangled in pending litigation concerning these multifaceted issues. This means that a definitive, universally accepted legal framework or "solution" to these problems is not imminent. Gellis predicts that the legal landscape will continue to evolve through an iterative process. "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," she stated. Despite the ongoing uncertainty and potential for conflicting rulings, these early decisions are undeniably shaping current practices. "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." The legal journey for AI and copyright is just beginning, with each court decision, even if provisional, setting precedents that guide the industry’s ethical and operational choices.

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.