Anthropic’s $1.5-billion settlement signals new era for AI and artists

The Chatbot Anthropic manufacturer has agreed to pay $ 1.5 billion to the authors in a historic copyright regulation which could redefine how artificial intelligence companies compensate the creators.

The startup based in San Francisco is ready to pay the authors and publishers to settle a legal action which accused the company of illegally using their work to form its chatbot.

Anthropic has developed an AI assistant named Claude who can generate text, images, code and more. Writers, artists and other creation professionals have raised fears that anthropogenic companies and other technologies use their work to train their AI systems without their permission and not compensate them fairly.

As part of the regulation, which the judge must still be approved, Anthropic agreed to pay $ 3,000 to the perpetrators per work for around 500,000 pounds. It is the largest regulation known for a copyright case, signaling to other technological companies faced with allegations of copyright violation that they could also pay rights holders.

Meta and Openai, the chatgpt manufacturer, were also prosecuted for alleged copyright violation. Walt Disney Co. and Universal Pictures continued the company Ai Midjourney, which, according to the studios, formed its models of image generation on their material protected by copyright.

“It will provide significant remuneration for each class work and will establish a precedent forcing AI companies to pay copyright holders,” said Justin Nelson, a lawyer in a statement. “These regulations send a powerful message to the companies and creators of the AI, that the taking of works protected by copyright on these pirate websites is wrong.”

Last year, the authors Andrea Bartz, Charles Graeber and Kirk Wallace Johnson continued Anthropic, alleging that the company had committed a “large -scale flight” and trained its chatbot on pirated copies of books protected by copyright.

The American district judge William Alsup of San Francisco judged in June that the use of books by Anthropic to form the models of AI constituted “fair use”, so it was not illegal. But the judge also judged that the startup had badly downloaded millions of books in online libraries.

The fair use is a legal doctrine in the American law on copyright which allows the limited use of material protected by copyright without authorization in certain cases, such as teaching, criticism and reports. IA companies stressed this doctrine as a defense when they pursued by pursuing alleged copyright violations.

Anthropic, founded by former Openai employees and supported by Amazon, was hacked at least 7 million books of books3, Library Genesis and Pirate Library Mirror, online libraries containing unauthorized copies of books protected by copyright, to form his software, according to the judge.

He also bought millions of copies printed in bulk and stripped the bindings of the books, cut their pages and scanned them in the form of a digital and readable by machine, which Alsup judged within the limits of fair use, according to the judge’s decision.

In a later order, Alsup highlighted the potential damage to copyright rights downloaded from the Shadow Libgen and Pilimi libraries by Anthropic.

Although the price was massive and unprecedented, it could have been much worse, according to some calculations. If Anthropic was billed a maximum penalty for each of the millions of works he used to form his AI, the bill could have been more than 1 dollars, suggest certain calculations.

Anthropic did not agree with the decision and did not admit the reprehensible acts.

“Today’s regulations, if approved, will resolve the inherited inherited complaints from the applicants,” said Aparna Sridhar, lawyer general of Anthropic, in a statement. “We remain determined to develop safe AI systems that help people and organizations to extend their capacities, advance scientific discovery and solve complex problems.”

The anthropogenic dispute with the authors is one of the many cases where artists and other content creators put the companies at the origin of the generating AI to compensate for the use of online content to train their AI systems.

The training consists in nourishing enormous amounts of data – including publications on social networks, photos, music, computer code, video and more – to form AI robots to discern the models of language, images, sound and conversation that they can imitate.

Certain technological companies have prevailed in copyright prosecution registered against them.

In June, a judge rejected a legal action filed against the Facebook Meta parent company, which also developed an AI assistant, alleging that the company had stolen its work to train its AI systems. The American district judge Vince Chhabria noted that the trial had been launched because the complainants “made the bad arguments”, but the decision “did not defend the proposal that meta use of copyright to form his linguistic models is legal”.

Trade groups representing publishers welcomed the anthropogenic regulations on Friday, noting that it sends a large signal to technological companies that develop powerful artificial intelligence tools.

“Beyond the monetary terms, the proposed regulations offer enormous value in sending the message that artificial intelligence companies cannot illegally acquire content in ghost libraries or other pirate sources as constituent elements of their models,” said Maria Pallant, President and Chief Executive Officer of the Association of American Publishers.

The Associated Press contributed to this report.