Major artificial intelligence companies including Meta, Anthropic, OpenAI, and others have won the right to use copyrighted materials for training AI models under the doctrine of fair use. These rulings, delivered by federal judges in the United States, are set to reshape the landscape of AI development, content creation, and digital copyright for years to come.
The most recent decisions came in high-profile lawsuits brought by authors and creators who accused AI companies of using their books and other works without permission to train large language models (LLMs). In one pivotal case, a judge ruled that Anthropic’s use of millions of books to train its Claude AI models qualifies as fair use, a legal doctrine that allows certain uses of copyrighted content without the owner's consent.
Similarly, Meta secured a victory when a federal judge found that training its Llama AI on copyrighted works was "transformative" enough to be considered fair use, dismissing the copyright claims brought by a group of authors.
The courts found that training AI models on copyrighted works is "quintessentially transformative" meaning the process creates something fundamentally different, rather than simply replicating or competing with the original content. Judge Alsup, in the Anthropic case, compared it to teaching schoolchildren to write: “This is not the kind of competitive or creative displacement that concerns the Copyright Act. The Act seeks to advance original works of authorship, not to protect authors against competition”
These rulings provide legal certainty for AI giants, allowing them to scrape, copy, and use vast amounts of online content to train their models without paying licensing fees or seeking permission. This is a significant win for companies like Google, Meta, OpenAI, Microsoft, and Anthropic, who rely on massive datasets to power their generative AI products.
The decisions are a setback for authors, musicians, artists, and news organizations who have argued that AI training on their works undermines their market and creative rights. While the courts acknowledged that AI-generated content could compete with original works, they found the plaintiffs’ arguments insufficient to block the practice under current law.
While these victories are decisive, they are not absolute. Judges noted that future plaintiffs could potentially win if they can prove that AI-generated outputs flood the market with competing works that harm the original creators’ livelihoods. Additionally, some issues such as the use of pirated copies for training remain unresolved and will be addressed in upcoming trials.