Andrea Bartz, Charles Graeber and Kirk Wallace Johnson were in the San Francisco courtroom to watch history: Judge Alsup approved the largest settlement in the history of US copryight law at $1.5 billion.
Afterwards, the book authors issued a statement:
We are grateful for the Court’s action today, which brings us one step closer to real accountability for Anthropic and puts all AI companies on notice they can’t shortcut the law or override creators’ rights.
Andrea Bartz, Charles Graeber and Kirk Wallace Johnson
Under the terms of the settlement, the copyright owner(s) for each book that falls within the class of 482,460 books downloaded from shadow libraries Library Genesis and Pirate Library Mirror is entitled to claim $3,000 from the class funds. If the copyright owner has more than one book in the class, they can get recovery for each book.
The default approach for situations involving shared interests between publisher and author is a 50/50 split. But “[a]ny claimant … can de-select that default by indicating that the claimant has a different contractual arrangement that it seeks to pursue. Id. ¶ 4(e). For education works, 2
more individualized determination of the proper allocation of the award will occur.”
The class criteria: “(1) works in the LibGen and PiLiMi downloaded by Anthropic that (2) have an “ISBN or ASIN which was (3) registered with the United States Copyright Office within five years of the work’s publication and which was registered with the United States Copyright Office before being downloaded by Anthropic, or within three months of publication.”
The anthropic settlement sets a $1.5 Billion benchmark for other cases
The Anthropic settlement now provides a $1.5 billion benchmark for copyright lawsuits against other AI companies, especially the proposed book author class actions against OpenAI, Microsoft, and Apple. If the judges in those cases adopt the same view of downloading and class certification as Judge Alsup, it’s hard to image that the book authors would be willing to settle for any amount below $1.5 billion.
And it will be interesting to see if any of the lawsuits involving other types of works — images, videos, and music — later settle. The plaintiffs will likely adopt the same strategy as Bartz in trying to treat the initial downloading of any unauthorized datasets as a separate act of copyright infringement, no matter if later used for training an AI model.
Related Stories