A US district judge has granted a request for class certification by journalists and book authors who claim their copyrighted works were pirated and used without their authorization to train the Anthropic AI tool, Claude.
As Judge Alsup’s introduction states,
This civil action exemplifies the classic litigation that should be certified as a representative action, for the entire class stands aggrieved by defendant’s downloading of their books from pirate libraries on the internet. It will be straightforward to prove the class-wide wrong done. In short order, counsel can prove defendant violated the Copyright Act by doing Napster-style downloading of millions of works. Defendant, however, objects that it will be a litigation nightmare to determine the specific works it downloaded and, as to each, who is the rightful victim of the infringement.
As the judge explained,
In early 2021, a co-founder of Anthropic downloaded 196,640 unauthorized copies of copyrighted books from the pirate library known as Books3…. Anthropic’s co-founder downloaded the pirated books to avoid the trouble of paying for them, hoping they might prove useful for training large language models (LLMs) or for something else.
Not satisfied with this initial supply, the co-founder “used the infamous BitTorrent protocol to copy books peer-to-peer from decentralized copies of another pirate library” – collecting another five million copies, and then another two million from another pirate source.
Later, Anthropic also purchased and scanned books.
But it turned out that
once Anthropic stopped using pirate-sourced copies of books to train LLMs, its models suffered a “performance hit.” It wanted to “recover” from this by finding books at least as good as the “old-books quality,” meaning the pirated ones.
Under Rule 23 of the Federal Rules of Civil Procedure, one or more members of a class may sue as representative parties on behalf of all members only if:
- the class is so numerous that joinder of all members is impracticable;
- there are questions of law or fact common to the class;
- the claims or defenses of the representative parties are typical of the claims or defenses of the class; and
- the representative parties will fairly and adequately protect the interests of the class.
Here, the judge approved the definition of the relevant class as (in part),
All beneficial or legal copyright owners of the exclusive right to reproduce copies of any book in the versions of LibGen or PiLiMi downloaded by Anthropic….
Anthropic had argued that authors and author-affiliated entities didn’t adequately represent owners of other copyrights because “academics, researchers, and writers who use [LLMs] disagree with the position taken by plaintiffs and actively use and benefit from LLMs like Claude.”
The judge found this unpersuasive, saying that
Whoever would like Anthropic to take their books for free can achieve that result even if the remaining class prevails: They can opt out and license their works to Anthropic or to all the world for nothing.
Copyright Office are entitled to receive statutory damages – which can be as much as $150,000 per title for willful infringement.
The judge concluded:
There is no serious prospect that these claims can be addressed through individual actions. A denial of a motion to certify the class would amount to a concession that copyright owners’ credible allegations of infringement will go unchecked by courts so long as a copyist allegedly violates the Copyright Act not a little but a whole lot. “[T]o do so would eliminate the class action deterrent for those who engage in complicated and imaginative rather than straightforward schemes” — or, in this instance, straightforward piracy but at massive scale.
Just like the haiku above, we like to keep our posts short and sweet. Hopefully, you found this bite-sized information helpful. If you would like more information, please do not hesitate to contact us here: https://aeonlaw.com/contact-us/.
