US courts assess AI training on copyrighted works differently — TechCrunch
US courts have not yet developed a unified approach to whether companies may train artificial intelligence models on books and other copyrighted works. In one of the first rulings of this kind, Judge William Alsup found Anthropic’s training of models on authors’ works lawful, but ordered the company to pay $1.5 billion as part of a settlement over claims concerning the acquisition of books from illegal shadow online libraries. TechCrunch reports.
The judge distinguished between the training of large language models itself and the way the company obtained the books. In his ruling, he compared a model’s processing of a vast body of texts to a person reading literature in an effort to create something new rather than reproduce or replace the originals.
The question of fair use
Intellectual property lawyer Cathy Gellis believes that this ruling is generally favorable to AI developers. According to her, the court viewed model training more as becoming acquainted with a work than as copying it. At the same time, copyright law has not been updated since 1976, so courts have to apply old rules to technologies that did not exist at the time.
In such cases, the principle of fair use is key. In certain circumstances, it permits the use of protected materials without the direct permission of the rights holder, including for criticism, parody, or education. When assessing cases, courts take into account the purpose and nature of the use, the amount of material used, and the effect on the market for the original work.
More current news is available on the UA.News Telegram channel Telegram.
Competition with the rights holder
Jason Henderson, senior lawyer and founder of the IP & Media practice at JWL International, noted that case law in AI-related cases is inconsistent. At the same time, courts tend to assess the use of others’ content more negatively when a new product directly competes with the rights holder.
This approach was applied in the case of Thomson Reuters v. Ross Intelligence. Judge Stephanos Bibas ruled that the use of Thomson Reuters content to create a competing AI-based legal platform was not transformative and did not qualify as fair use. The authors’ argument that chatbots may compete with them by generating synthetic books has not yet received support in court.
The issue of copyright for AI-generated content remains separate. In Thaler v. Perlmutter, the court ruled that a work generated entirely by artificial intelligence cannot be subject to copyright. Most lawsuits against AI companies in this area are still pending.
