On September 5, Anthropic, the AI startup behind the Claude chatbot, announced a jaw-dropping $1.5B settlement in a class-action copyright lawsuit brought by a coalition of authors and publishers. The accusation? The writers asserted that the pirated copies of their books fueled AI training data.
Let’s take a deep dive into what happened, why it matters, and how this could potentially influence the future of AI and copyright law.
A Brief Recap: What Happened?
Last year, a group of authors filed a class-action lawsuit against Anthropic for the illegal use of their books in training the Claude chatbot. Among the authors were notable figures like Andrea Bartz, a journalist and NYT bestselling author; Charles Graeber, author of The Good Nurse; and Kirk W. Johnson, author of The Feather Thief.
The authors didn’t just allege that Anthropic used their copyrighted works. They specifically claimed the pirated works were obtained from shadow libraries like LinGen and Pirate Library. This fact served as a crucial distinction in Anthropic’s undoing.
Now, over the years, AI companies have been training their LLMs using vast amounts of text data scraped from the internet. Source materials used for this purpose range from books, articles, and other copyrighted works.
While AI developers claim fair use, many authors, journalists, and publishers argue it’s outright theft.
When the authors brought the case before Judge William Alsup in San Francisco federal court, he ruled that Anthropic’s use of their books complied with US law.
In other words, if developers legally obtain materials to train AI models, courts deemed it “fair use.” However, the court could not dismiss the case because Anthropic illegally sourced the materials.
Due to this pivotal point in the case, Anthropic chose to settle, although they didn’t admit to any wrongdoing. The case involved a total of over 500,000 books and a payout of $3,000 per book, making it the largest payout in the history of US copyright cases.
Why This Settlement Is a Big Deal
This copyright case has set a precedent for how similar cases will be managed. Anthropic is not the first company to be sued over copyright infringement, and it certainly won’t be the last.
When we take a look at the cost of the settlement alone, $1.5B isn’t exactly chump change; that’s almost 10% of Anthropic’s market cap. If settlements like this stick, the cost of training AI models will skyrocket.
AI models need massive data sets to function at human thinking capacity. “Free data scraping” streamlined this process and cut costs, largely enabled by abundant free online content.
Now, authors will have the advantage of copyrighting their works and selling licensing fees at much higher prices, as companies would rather not go through the purse-breaking process of a settlement.
It’s practically a double-edged sword. For big tech it could be a huge inconvenience due to delays and limited access to data. And for writers and publishers, it could signify a new era where their works are highly appreciated.
The ripple effect, however, doesn’t end here. Outside the US, Europe’s tightening AI Act already mandates data transparency. Currently, the US is lagging behind in the implementation of laws regarding AI copyright infringement, and this issue with Anthropic could well be the wake-up call to do so.
What’s Next?
Another hearing looms on the 25th of September, where Judge Alsup will revisit the settlement. Should the judge be satisfied with the clarified terms, the approval for the settlement could be granted.
However, the ripple effect caused by this case would be far from over even if the settlement is approved. Anthropic’s $1.5B settlement will serve as a global reference point, a signal that creators’ rights must be part of AI’s future.
