Artificial‑intelligence firms that power chatbots such as ChatGPT, Gemini and Claude have built their models on vast corpora of text, including millions of books, articles and academic papers. Most of those works were harvested without authors’ permission, prompting a wave of lawsuits that question whether the practice violates copyright law.
In a landmark case last year, U.S. District Judge William Alsup ordered Anthropic, the startup behind Claude, to pay $1.5 billion to a group of writers whose books were used to train the company’s language model. The judge’s opinion, however, stopped short of declaring the training itself illegal. Instead, he ruled that Anthropic’s liability stemmed from pirating the texts from illegal online shadow libraries, not from the act of training a model on copyrighted material.
Attorney Cathy Gellis, who specializes in intellectual‑property and technology law, sees the ruling as a win for AI companies. "The decision treats reading a copyrighted work as different from copying it," she told TechCrunch. "Copyright law hinges on copying, not on simply consuming the work." Gellis added that a $1.5 billion fine is modest for a firm that projects $200 billion in annual revenue by 2028.
Legal scholars point to fair‑use doctrine as the crux of the debate. Courts evaluate purpose, nature, amount used and market impact to decide if a use is transformative enough to be exempt. In the Thomson Reuters v. Ross Intelligence case, a judge concluded that training an AI on Reuters’ content to launch a competing legal‑research platform was not transformative and therefore not fair use.
Other rulings add layers of complexity. In Thaler v. Perlmutter, the court held that a work generated entirely by AI cannot be copyrighted, raising questions about ownership of AI‑produced content and how much human input is required for protection.
Because the United States’ copyright statutes have not been revised since 1976, judges must interpret outdated language for modern technology. "Everybody is very worried because the law is all over the place," said Jason Henderson, senior attorney at JWL International. "The courts are still figuring out how to apply fair‑use principles to AI training."
For now, AI developers continue to face litigation while navigating a patchwork of legal opinions. The industry watches each decision closely, knowing that future rulings could either cement a permissive stance on data scraping or impose stricter limits that reshape how models are built.
Until Congress updates the statute or a definitive Supreme Court ruling emerges, the boundary between lawful training and copyright infringement will remain blurred, leaving both creators and AI firms in a state of uncertainty.
Este artigo foi escrito com a assistência de IA.
News Factory APP - notícias agênticas para impulsionar seu SEO e AEO.