The author agrees to pay $ 1.5B to settle the author class action on anthropic AI training
Anthropic on Friday told a federal judge of San Francisco that he had agreed to pay $ 1.5 billion to settle a class-action case from a group of authors who accused the Artificial Intelligence Company of using pirated copies of their books, which to train their AI Chatbott, Cloud without permission.
In a court, Anthropic and Vadi asked US District Judge William Alasup to approve the settlement, after announcing the agreement in August without disclosing the conditions or funds.
The plaintiff said in the filing, “If approved, it will be a publicly reported copyright recovery in the landmark settlement history, which is motivated for final decision compared to any other copyright class action settlement or any individual copyright case.”
The proposed deal marks the first settlement in a string of cases against technical companies including OpenIA, Microsoft and Meta platforms, which is on the use of copyright material to train the copyright AI system.
As part of the settlement, Anthropic stated that it would destroy the copies of books obtained through pirateing sites Libgain and Pilimi (Pirate Library Mirror). Under the deal, it can still face claims of violations related to material produced by the company’s AI model.
In a statement, Anthropic stated that the company is committed to developing “safe AI systems that help people and organizations expand their abilities, pursue scientific discovery and solve complex problems.” The agreement does not include the entry of liability.
“This historic settlement is an important step in accepting that AI companies cannot only steal the creative tasks of authors to create their AI, as they require books to develop quality LLM,” the author said the guild’s CEO Mary Rasenburger. statement,
“These huge rich companies, at the cost of billions, steal from those who barely earning an average income of $ 20,000 (US) in a year. This disposal sends a clear message that AI companies will have to pay for books that they pay for other essential components of their LLM.”
Although the estimated seven million books were downloaded by anthropic from piracy sites, according to the author Guild, only 500,000 works are covered in class action, which means the settlement amount is about $ 3,000 per writer.
Writers Andrea Bartz, Charles Grabber and Kirkas Johnson filed a class action against Anthropic last year. He argued that the company, which is supported by Amazon and the alphabet, illegally used millions of pirated books to teach its AI auxiliary cloud to respond to human signals.
Creative work was stolen
The allegations from the authors resonated dozens of other cases brought by writers, news outlets, visual artists and others, saying that tech companies stole their work to use in AI training.
Companies have argued that their systems use copyright content properly to create new, transformative materials.
Alsup ruled in June that Anthropic properly used books to train Cloud, but found that the company violated its rights by saving more than seven million pirated books in a “Central Library”, which is not necessarily used for that purpose.
One of the best-selling Vancouver writer has launched a class-action case against Nvidia, Meta and two other technical giants. JB McInon claims that the books he and other Canadian writers were used to train the artificial intelligence models illegally.
In December, a test was initiated to determine how much anthropic outstanding for alleged piracy, with a possible loss in hundreds of billions of dollars.
Other AI copyright cases are still being debated on the decisive fair-use question.
Vancouver author JB McKinon recently launched class-action cases against Nvidia, Meta, Anthropic and Databricics Inc. in the BC Supreme Court, alleging that the works of her and other Canadian writers have been illegally used for AI training.
Another San Francisco judge, hearing the uniform ongoing trial against Meta, ruled immediately after Alsup’s decision that using copyright work without permission to train AI would be illegal in many circumstances. “
