Federal Judge Rules Meta’s AI Training on Books is Fair Use
Created on June 26|Last edited on June 26
Comment
A federal judge in California has ruled in favor of Meta in a closely watched copyright lawsuit involving the use of books to train artificial intelligence models. Thirteen authors, including Sarah Silverman, sued the company alleging that Meta illegally used their copyrighted books to train large language models. Judge Vince Chhabria issued a summary judgment, meaning he ruled without a jury trial, determining that Meta’s actions fell under the legal protection of “fair use.”
Fair Use and the Limits of the Decision
Judge Chhabria’s decision focused on the argument that Meta’s use of the books was transformative, meaning the AI models did not simply regurgitate the texts but instead used them in a way that added new meaning or utility. He also found that the plaintiffs failed to show meaningful evidence of market harm, which is one of the key criteria for establishing copyright infringement. Importantly, the judge cautioned that this ruling does not give companies a blanket legal cover for using copyrighted content in AI training, saying the case was decided based on the specific facts and the weakness of the plaintiffs’ legal strategy.
Parallel Rulings and Industry Implications
This ruling follows a similar decision in favor of Anthropic, another AI firm facing allegations of unauthorized use of copyrighted materials. Both cases are being closely followed by the tech and publishing industries because of their implications for how AI models are developed and trained. While the decisions have favored the tech companies so far, judges have emphasized that future cases could go differently if plaintiffs bring more robust evidence or better arguments.
Market Harm and Legal Weakness
A central weakness in the plaintiffs’ case, according to the judge, was the absence of solid evidence showing that Meta’s use of the books diluted their market value. Without a clear showing that Meta’s models substituted for the original works or diminished their economic potential, the court could not conclude that infringement occurred. Judge Chhabria was direct in stating that the plaintiffs “made the wrong arguments” and failed to develop the right ones with sufficient evidence.
Other Active Lawsuits and What’s Next
The ruling comes as other lawsuits are still pending, including a high-profile one from The New York Times against OpenAI and Microsoft over the use of news articles for model training. Entertainment giants like Disney and Universal are also suing Midjourney over AI training involving film and television content. Judge Chhabria’s opinion hinted that the impact of AI on different markets—like journalism—might be more direct, and thus more legally vulnerable, than it is for books.
Context for the Broader AI Landscape
These rulings mark a cautious but notable legal validation for companies like Meta and Anthropic, but they don’t close the book on the issue. They instead underline how much hinges on specific facts: what kind of material is used, how it’s used, and how well either side can prove harm or transformation. Courts are not yet offering sweeping precedents that apply across the AI sector. Each case is likely to hinge on the details.
For now, Meta has secured a legal win, but the larger battle over copyright and AI training is still in its early stages.
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.