Federal judge: Anthropic acted legally with AI Book Training | Businessman

The federal judge for the first time decided to be legal for $ 61.5 billion AI, anthropic, to train his model AI in Copyrightd books without compensating or attributing the authors.

The US district judge William Alsup of San Francisco said in the judgment filed on Monday that Anthropion’s use of copyright, published books to train his model AI was “fair use” under the US copyright law was “exceptionally transformative”. Alsup compared the situation with a human reader who learns how to be a writer of books to create a new work.

“Like any reader who seeks to be a writer, Anthropic’s (AI) trained in works to exist and replicate or replace – but turned the hard corner and created something else,” Alsup wrote.

According to the decision, although Anthropion’s use of books protected by copyright as training material for Claude was fair, the court will organize a court on pirate books used to create the Central Library of the Anthropic and determine the resulting damage.

Related: “Extraordinary expensive”: Getty Images pours millions of dollars into one AI action, says CEO

The decision, the first time the federal judge with technology companies over creatives in the copyright lawsuit, creates previous shorts that prefer AI companies over individuals in copyright AI disputes.

These copyright courts rely on how the judge interprets the doctrine of fair use, the concept of copyright, which allows copyrights to use without obtaining permission from the copyright holder. The decision of fair use depends on how different work is from the original, what the end work is used for and whether it is replicated for commercial profit.

The petitioners in the case of class events, Andrea Bartz, Charles Graber and Kirk Wallace Johnson, are all authors who claim that Anthropic has used his work to train his chatbot without their permission. In August 2024, they filed an initial complaint, Bartz v. Anthropic, and claimed that anthropic sought the copyright law of piracy of books and replicated them to train its AI Chatbot.

The decision states in detail that anthropic has downloaded millions of books protected by free copyrights from pirate websites. Starting also Bough printing a copy of books protected by copyrights, some of which already had in their pirate library. Employees to link these books, reduce pages, scanned and saved them in digital files to add to the central digital library.

From this Central Library of Anthropic, she selected various groups of digitized books to train his AI Chatbot, Claude, the primary driver of the company.

Related: “A bottomless pit of plagiarism”: Disney, universal ensemble first Hollywood court against AI startup

The judge ruled that because Claude’s output was “transformative”, anthropic was allowed to use authorial work in the doctrine of fair use. Anthropic, however, still has to go to court over the books that pirate.

“Anthropic was not entitled to use pirate copies for its central library,” he reads the government.

Claude proved to be lucrative. According to the decision, anthropic earned anthropic annual income from corporate customers and individuals who paid for regulations for using AI Chatbot last year. Payed subscription for Claude ranges from $ 20 per individual up to $ 100 per one.

Anthropic faces another court from Reddit. In a complaint filed at the beginning of this month in a court in northern California Reddit, he claimed that anthropic was using his training material for AI without permission.