Recently, major technology companies, Anthropic and Meta each secured landmark victories in separate copyright lawsuits. The companies had been sued by authors and their publishers, regarding claims that these companies’ AI models were trained on copyrighted materials without claimant’s permission, in violation of U.S. copyright law. These are the first major decisions in a series of similar copyright suits and could signal a potential significant impact on the AI industry. The decisions are generally good for the industry, but the details still leave room for plaintiffs to succeed.
To train large language models, AI companies have collected extensive amounts of data and fed them into their systems. Generally, the more data a model is trained on, the more accurate its responses become. This data includes movies, articles, social media posts, and — central to the lawsuits in question — books. The data is often copyrighted, yet several companies have used it without obtaining permission from the authors or paying licensing fees. Companies have argued that this training process is considered “fair use.” Other major AI players, including Google, OpenAI, and Microsoft are involved in similar litigation. The plaintiffs range from independent authors to major entities like Getty Images and The New York Times.
These outcomes are set to have a big impact on the AI industry — determining if companies can continue using such data without paying a speeding ticket. If courts rule against them, companies will need to start paying for the training data with licensing deals. Combined with the vast amounts of training data being used, these licensing deals could cripple an industry that is already losing money.
In Anthropic’s case, Senior Judge William Alsup ruled that using copyrighted books to train its AI model, Claude, was transformative and therefore fair use. In his 32-page opinion, he likened the training process to teaching a human how to read and write, noting that readers are not obligated to pay each time they re-read a book or recall the information they learned. He also held that using the works to train an AI system is “among the most transformative [technologies] many of us will see in our lifetimes.” However, Judge Alsup held that the library of pirated content used in the training process was illegal. Anthropic will face a separate trial over the piracy claims.
In Meta’s case, District Judge Vince Chhabria took a different approach, instead focusing on how AI-generated works will impact the market for human-created content. For example, the market for biographies of public figures could be harmed if AI systems can generate endless variations of the same story. In contrast, the market for autobiographies would likely remain unaffected because readers value the human author’s perspective. Judge Chhabria ruled in favor of Meta, although emphasizing the highly fact-dependent nature of his decision. Judge Chhabria highlighted that the ruling only applied to the thirteen authors in the case and does not broadly legalize Meta’s use of copyrighted materials to train its AI systems.
Both rulings are likely to influence similar lawsuits moving forward, but they are only the beginning. Both sides still have compelling legal arguments, and the judges’ emphasis on fact-specific analysis suggest that outcomes will vary case-by-case. With other ongoing lawsuits and the likely appeals to follow, there is still a long way to go before this issue is settled.
For more information regarding the Anthropic or Meta disputes, please visit following:
[View source.]
