"Anthropic admitted to having bought millions of physical books and then digitizing them. The company also downloaded millions of pirated books from the notorious Library Genesis and Pirate Library Mirror troves of stolen material. The company nonetheless won part of the case, on grounds that scanning books is fair use and using them to create "transformative works" - the output of an LLM that doesn't necessarily include excerpts from the books - was also OK."
"Plaintiffs intended to pursue court action over those pirated works, but the filing details a proposed settlement under which Anthropic will create a $1.5 billion fund which values each pirated book it used for training at $3,000. Anthropic also agreed to destroy the pirated works. In the filing, counsel observes that this is the largest ever copyright recovery claim to succeed in the USA and suggest it "will set a precedent of AI companies paying for their use of alleged pirated websites.""
Anthropic admitted purchasing and digitizing millions of physical books and downloading millions of pirated books from Library Genesis and Pirate Library Mirror. A court found scanning books and using them to create transformative LLM outputs can qualify as fair use, but found that Anthropic broke the law by knowingly ingesting pirated books. Anthropic agreed to create a $1.5 billion fund valuing each pirated book used for training at $3,000 and to destroy the pirated copies. Counsel described the recovery as the largest copyright claim to succeed in the USA and said it will set a precedent for AI companies paying for alleged pirated website use. The filing asks the court to approve the settlement.
Read at Theregister
Unable to calculate read time
Collection
[
|
...
]