It's just the US, and the ruling has no direct implications for the Netherlands, but it does set a precedent for Dutch writers and other creatives. In a major lawsuit by writers against Anthropic, the 'nice' creator of AI system Claude, the court ruled that this tech giant may allow its system to learn by scanning books. The court finds that this falls under 'fair use'. So you could see that as a huge victory for the tech giants, but the judge did build in a huge 'but': Anthropic must have bought those books first.<\/p>\n\n\n\n
So the case revolves around how AI acquires the books. In this case, Anthropic was found to have drawn the books from an illegal database and that is not allowed. The court stated that learning from books is a human right, but stealing books is criminal. So Anthropic must first pay properly for all those books. After that, the company may do what it wants with them.<\/p>\n\n\n\n
Since millions of books and articles are involved, this could therefore amount to billions of euros for the AI guys. Whether writers will benefit much is questionable. After all. AI may buy your book once, but then they share the knowledge gained with millions of users. Too bad, of course, but the court ruled - by US standards, apparently rightly so - that there is no difference between a very popular human user with millions of followers, using and applying your knowledge, and a computer programme doing the same.<\/p>\n\n\n\n
What this might mean for Dutch copyrighted work is still unclear. We do not have the 'fair use' principle here, but the law does allow for certain exceptions. So the battle is not yet over, but one could imagine that such an AI would have to subscribe to Culture Press before they could pull the site dry.<\/p>\n\n\n\n