Anthropic, the company behind the Claude large language models, has reached a $1.5 billion settlement with a group of authors for accessing their books illegally. Its settlement is extremely notable as the court filing states it’s the largest publicly reported copyright recovery in history. As part of the settlement, Anthropic has agreed to pay $3,000 per book, plus interest. It has also agreed to destroy the datasets that contained the allegedly pirated material.
The settlement, which still needs to be approved, was brought by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson. The lawsuit was filed in the US District Court for the Northern District of California.
The lawsuit brought by the authors accuses Anthropic of engaging in large-scale copyright infringement for allegedly using books from pirated datasets. A judge previously ruled in June that Anthropic’s use of books for training AI models was fair use, so it’s interesting to see Anthropic offer a settlement at all.
Despite his decision that the books were fair use, the judge ordered a trial take place in December to decide whether the company infringed on copyright by obtaining the works from the databases Library Genesis and Pirate Library Mirror. By settling, Anthropic can avoid this with potentially more favorable terms.
Anthropic has been able to pay the $1.5 billion settlement because it has just closed at $13 billion funding round which valued the company at a whopping $183 billion. The financing was led by Iconiq, Fidelity Management, and Lightspeed Venture Partners. These financing rounds are also good for new firms like Anthropic because they can work on products without needing to be financially viable yet, although that is the ultimate aim.
This settlement will be a warning to other AI companies and creators that are alleged to have been taking copyrighted works from pirate websites. Most of the other companies developing a large language model are suspected of having used pirated materials at one point or another, but have largely brushed off the practice. Now that these authors have successfully settled with Anthropic, they could go after other firms and other authors could try to win money from those too.
Overall, this is an important case as it helps to define what constitutes copyright infringement in this relatively new AI landscape and sets a precedence for future issues.
Source: CNBC | Image via Depositphotos.com