Sony Music Publishing, Warner Chappell Music, and entities linked to these companies filed a lawsuit against artificial intelligence (AI) company Anthropic on Aug. 28, alleging that the tech enterprise stole their copyrighted music material to train AI.
“Defendants Anthropic and its founders Dario Amodei and Benjamin Mann have conducted a brazen campaign of illegally torrenting, scraping, and downloading copyrighted works on a massive scale in order to develop, operate, and reap enormous profits from Anthropic’s ‘Claude’ series of artificial intelligence (‘AI’) models,” according to the Aug. 28 lawsuit, filed with the U.S. District Court for the Northern District of California, San Jose Division.
The lawsuit alleges that in June 2021, at the direction of Amodei and on behalf of Anthropic, Mann used BitTorrent to illegally download at least 5 million copies of pirated books from the Library Genesis pirate website. In July 2022, defendants also torrented at least another 2 million pirated books from the Pirate Library Mirror, the complaint accused.
Among these books were those containing lyrics and sheet music to hundreds or more copyrighted musical compositions owned by the plaintiff music publishers, the complaint said, adding that both Amodei and Mann also directed Anthropic employees to make the illegal downloads. Without their instruction and approval, “none of the foregoing torrenting activity would have occurred,” according to the lawsuit.
Defendants’ torrenting activity is legally “indefensible,” and each instance in which they torrented the plaintiffs’ works amounts to a “separate and standalone” copyright infringement that harms the music publishers, the lawsuit said.
In addition to these downloads, Anthropic allegedly harvested copies of various works owned by the plaintiffs by scraping the internet. They are accused of specifically opting to harvest unauthorized copies of the plaintiffs’ works so that they wouldn’t have to pay copyright holders a licensing fee to use such content for training the Claude AI models.
Related Stories
Moreover, the AI models generate copies of the pirated works when they output answers to users. Anthropic has taught Claude models to generate vast quantities of new AI song lyrics that compete with plaintiffs’ copyrighted works as “market substitutes,” the complaint said.
The lawsuit accused Anthropic of depriving music publishers of their licensing fees and undermining the existing and potential licensing markets for copyrighted lyrics.
It is crucial to ensure that AI development “occurs responsibly and on terms that are agreeable to rightsholders and in a manner that protects their rights, livelihoods, and the creative industries as a whole,” the lawsuit said. “Even the most revolutionary of technologies must develop within the bounds of the law, and Anthropic’s Claude models are no different.” The plaintiffs are seeking up to $150,000 in statutory damages per infringed work.
The Epoch Times reached out to Anthropic for comment but did not receive a response by publication time.
‘Transformative’ Use of Music
Anthropic argued that its use of plaintiffs’ lyrics to train AI is “transformative,” in the sense that the copyrighted music is part of a multi-trillion-token dataset to train an AI model about the world.
This is “the very definition of ‘transformative’ under the fair use doctrine,” the company argued. “The lyrics are literally transformed, in that they are broken down into small tokens used to derive statistical weights, rather than stored as intact copies.” The case is ongoing.
Other Lawsuits
The lawsuit alleged that Anthropic used pirated copies of their books to train Claude. If Anthropic had lost the case at trial, the judge could have imposed a penalty of up to $150,000 per violation.
Given the judge observed that more than 480,000 works were part of the claim, the potential award could have been in the tens of billions of dollars.
On Aug. 17, independent music publisher Round Hill Music sued Anthropic in the U.S. District Court for the Northern District of California over copyright violation, providing the court with a non-exhaustive list of 500 musical compositions that the company has allegedly infringed.
