Judge OKs Anthropic’s Use of Copyrighted Books in AI Training. That’s Bad News for Creators

The use by anthropic of books protected by copyright in its AI training process has been “extremely transformative” and fair use, the American senior district judge William Alsup judge Monday. This is the first time that a judge has decided in favor of an AI company on the question of fair use, in a significant victory for generative AI companies and a blow for creators.

Fair use is a doctrine that is part of the American copyright law. It is a four -part test which, When the criteria are metAllows people and businesses using protected content without the authorization of the rights holder for specific purposes, as when drafting a term document. Technological companies say that the exceptions for fair use are essential for them to access the massive quantities of content generated by the man they need to develop the most advanced AI systems.

Writers, actors and many other types of creators were just as clear By arguing that using their content to propel AI is not fair use. Publishers, artists and owners of content catalogs have brought prosecution alleging that IA companies such as Openai, Meta and Media infract their protected intellectual property to try to circumvent expensive but standard license procedures.

(Disclosure: Ziff Davis, CNET’s parent company in April, filed a complaint against Openai, alleging that it has violated Ziff Davis Copyrights in the training and exploitation of its AI systems.)

The authors who pursue Anthropic for copyright violation say that their books have also been illegally obtained-that is, they were hacked. This leads to the second part of Alsup’s decision, based on his concerns about anthropic methods to obtain books. In the decision, he writes that the anthropogenic co-founder Ben Mann knowingly downloaded unauthorized copies of 5 million pounds from Libgen and an additional 2 million of Pirate Library Mirror (Pirlimi).

The decision also describes how anthropic has deliberately obtained printed copies of the books he previously hacked in order to create “his own catalog of bibliographic metadata”. The anthropogenic vice-president, Tom Turvey, says the decision, was “responsible for obtaining” all the books in the world “while avoiding so much” legal / practical / business slog “.” This meant buying physical books from publishers to create a digital database. The anthropogenic team has destroyed and threw millions of second -hand books in this process; To prepare them for a scan readable by machine, they eliminated them from their bindings and cut them to adapt.

The acquisition and digitization of anthropic of printed books were fair use, says the decision. But that adds: “The creation of a permanent library for general use was not itself an equitable use of anthropic hacking.” Alsup has ordered a new trial concerning the pirated library.

Anthropic is one of the many companies of AI faced with complaints in matters of copyright in court, so this week’s decision is likely to have massive training effects in the industry. We will have to see how the hacking is resolved before knowing how much money anthropic can be ordered to pay for damages. But if the scales to grant several equitable exceptions to AI companies, the creative industry and the people who work there will certainly have damage.

For more, consult Our guide to understand copyright in the AI ​​era.