Dozens of similar copyright cases are now working through the courts, cases have been filed against all top players – not only Anthupak and Meta but also Google, Openi, Microsoft and more. On the other hand, the plaintiffs are from individual artists and authors to big companies such as Gati and New York Times.
The results of these cases are ready to have a lot of impact on the future of AI. In fact, they will decide whether the model makers can continue the free lunch order. If not, they will need to start paying such training data through new types of new types of deals – or find new ways to train your models. They can increase the possibility of industry.
And that is why last week’s victory is important for technology companies. So: Close cases? Not enough. If you drill the details, decisions are less cut and dry than ever. Let’s look closely.
In both cases, a group of authors (Anthropic Suite Class action was action. 13 Playmakers tried Meta, including high -profile names like Sara Silverman and Ta Niyasi Coats) to prove that a technology company used its books to train large language models. And in both cases, the companies argued that the training process is considered as a fair use, it is a legal supply that allows the use of copyright works for some purposes.
The similarities there end. Deciding in favor of Anthropic, Senior District Judge William al -Soph argued that the firm’s use of books was legal because he did what he did to them, it means that he did not replace the original actions but made something new to them. In his judgment, Alsop wrote, “There was the most changing technology that many of us would see in our lives.”
In the case of Meta, District Judge Vince Chibraia gave a different argument. He also joined the technology company, but he focused on his decision on the issue whether Meta had damaged the market for the authors’ work. Chbia said he believed that Alsop had made the importance of the damage to the market. He wrote on June 25, “The key question in any matter, where the defendant has copied one’s original work without permission, is whether allowing people to engage in such behavior will reduce the market substantially for the original.”
The same result; Two very different rules. And it’s unclear what it means for other matters. On the one hand, it reinforces at least two versions of a fair use argument. On the other hand, there is some disagreement on how to decide on a fair use.
But there are more things to note. Chbia was very clear in her decision that Meta did not win because it was on the right, but because the plaintiffs failed to make a considerable reason. He wrote, “In the great scheme of things, the results of this decision are limited.” “This is not a class process, so this decision only affects the rights of these 13 authors – not countless others whose work used to train their models for training. And, as it should be clear now, this decision does not decide that the use of copyright content for the training of the meta -language model.” It reads a lot like someone else’s complaint with complaints and another go.