A large group of authors and publishers whose work was illegally used by the AI company Anthropic achieved a victory last week when the two sides agreed to a payment of $1.5 billion to rectify the wrong. The agreement followed a district court judge’s ruling in June that Anthropic could use the writers’ material for AI training, as this could be considered a transformative use of their work, but that it had to acquire the copyrighted material legally.
The out-of-court agreement was especially driven by the substantial financial penalty that a company faces when it knowingly acquires copyrighted content, as Anthropic did.
The agreement was reported in The New York Times' Sept. 6 issue. While encouraging, it has no legal precedent, as the Times reported, because it was not determined by a court. And, it has not yet been affirmed by the judge. Included is the requirement that Anthropic delete what it downloaded.
Writers and publishers have long seen their material taken without payment by digital industries that repackage it for sale to their users. Music and movie artists and producers faced the same challenges two decades ago when their content was taken and made available without their permission; after a short period, courts halted that.
While $1.5 billion looks to be a significant amount – and if affirmed it is the largest payment for wrongly taken copyrighted material – critics point out it amounts to only about $3,000 for each of the 500,000 content pieces in the suit, and that it’s coming from a company that has acquired investor participation in amounts anywhere from $4 billion to about $18 billion, which according to news reports has totaled $100 billion or more.
While the Anthropic agreement is significant in more than dollars, there are numerous other aggregating companies, especially now with the advent of AI, that are taking copyrighted content from various sources. In one example, The New York Times agreed in May to license its content with Amazon to expand its availability, while earlier suing Open-AI for helping itself to what was described in a May 30 story as millions of Times articles to be used for AI training.
A large amount of content goes into making that robo resource as nuanced-filled as it can be, and that comes from the skills and efforts of creators, small and large. With the Anthropic agreement, the scales have tipped a bit in favor of the providers.
