Humanity told a federal judge in San Francisco on Friday that it agreed to pay $1.5 billion to resolve a class action lawsuit from a group of authors who accused the artificial intelligence company of using copies of the book to train Claude, an AI chatbot.
The plaintiffs in the Humanity and Court applications asked District Judge William Alsup to approve the settlement after announcing the contract without disclosing the terms or amount.
“If approved, this landmark settlement will be the largest publicly reported restoration of copyright in history, greater than individual copyright cases litigated in other copyright class action settlements or final judgments,” the plaintiff said in his filing.
The proposed transaction marks the first settlement of a series of lawsuits against high-tech companies, including Openai, Microsoft and the Meta platform, over the use of copyrighted materials to train generative AI systems.
As part of the settlement, Anthropic said it would destroy downloaded copies of the books it obtained through the Pirating sites Libgen and Pilimi (Pirate Library Mirror). Under the agreement, you may face infringement claims related to materials generated by the company's AI model.
In a statement, Humanity said the company is “working on developing safe AI systems that help people and organizations expand their capabilities, advance scientific discoveries and solve complex problems.” The contract does not include approval of liability.
“This historic settlement is an important step in acknowledging that AI companies cannot simply steal the creative work of authors to build AI just because they need a book to develop quality LLM.”
“These very rich companies stole from people who are worth billions and earn a median income of just $20,000. [US] year. The settlement sends a clear message that AI companies must pay for the books they use, just as they pay for other important components of LLM. ”
According to the Author Guild, an estimated 7 million books have been downloaded by humanity from copyright infringing sites, meaning that a settlement amounts to around USD 3,000 per author.
Authors Andrea Burtz, Charles Graeber and Kirk Wallace Johnson filed a class action lawsuit against humanity last year. They argued that the company, backed by Amazon and Alphabet, would illegally use millions of pirated copies to teach AI assistant Claude to respond to human prompts.
Stolen creative work
The writer's allegations reflect dozens of other lawsuits that say tech companies, including authors, news outlets and visual artists, have stole jobs they use in AI training.
Companies argue that the system will use copyrighted materials fairly to create new, transformative content.
In June, Allsup determined that humanity had used the book fairly to train Claude, but the company found that it had violated its rights by saving more than seven million pirated copies in a “central library” that is not necessarily used for that purpose.
The bestseller Vancouver author has launched a class action lawsuit against Nvidia, Meta and two other tech giants. JB Mackinnon claims that books written by him and other Canadian authors were illegally used to train artificial intelligence models.
The trial was scheduled to begin in December to determine that there was a suspected copyright infringement, with potential damages potentially reaching hundreds of millions of dollars.
Important fair use questions are still discussed in other AI copyright cases.
Vancouver author JB Mackinnon recently launched a class action lawsuit against Nvidia, Meta, Anthropic and Databricks Inc. in the BC Supreme Court.
Another San Francisco judge heard a similar ongoing lawsuit against Meta shortly after Allsup's decision that using copyrighted work without permission to train AI is illegal in “many circumstances.”
