The first U.S. court decision on AI fair use is not good
Image credit: Bermix Studio
A landmark ruling between copyright holders and AI training models has concluded in a case of fair use, and it favoured AI.
The creative industries look on as the use of AI proliferates. The industry represents a threat on multiple fronts: It dilutes human-created art but it also trains itself on the work of human creators to understand how to reproduce creative works.
The development of AI models is moving so quickly that governments and legal protections have been slow to respond. The UK, for instance, has been attempting to change the law surrounding copyright for AI to the pushback of music companies over the outcome.
This month has seen the first court decision in the U.S. over AI training. The case saw a group of authors take AI company Anthropic to task over the use of their copyrighted content to train their AI model.
The complaint stated that Anthropic had used millions of copyrighted books to train its AI model ‘Claude’. The case sits among many in which creators and copyright owners go up against AI companies for using music, books, paintings, and more to train their models. This ruling has become the first conclusion.
Unfortunately, the judge ruled that Anthropic’s unconsensual use of copyrighted material to train an AI model that generates material based on that content comes under fair use if it has been legally purchased. The judge ruled that the AI’s use of the copyrighted content was “spectacularly transformative” in its output.
However, because Anthropic had pirated access to millions of the books it had used to train their AI model, a copyright infringement case will go forward. The judge ruled that whilst training their model on purchased content was fair use, “Anthropic had no entitlement to use pirated copies”.
Whilst the ruling was somewhat in favour of the AI company – and could set a precedent for other such cases – the US copyright law allows copyright owners to claim up to $150,000 per infringed material. This could see Anthropic face substantial damages, despite their success in the case.
There is also a chance that similar cases against the use of copyrighted content to train AI see different conclusions. That may see the case against Anthropic changed if the creators decide to appeal the ruling, which they most likely will.
Copyright lawyer John Strand commented on the precedent of the case, saying: “There are dozens of other cases involving similar questions of copyright infringement and fair use pending throughout the U.S., and Judge Alsup’s decision here will be something those other courts must consider in their own case.”