Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
That is an exception UMG vs Anthropologie Case in point, because at least early on, earlier versions of Anthropic would output song-by-song lyrics. This is a problem. The current status of that case is that they have safeguards in place to prevent that from happening, and the parties have agreed that, pending resolution of the case, those safeguards are sufficient, so they are no longer seeking a preliminary injunction.
At the end of the day, AI is not a difficult question for companies Is it legal to engage in training? it is What do you do when your AI produces output that is very similar to a specific task?
Do you expect most of these cases to go to trial, or do you see settlements on the horizon?
There may be some settlements there. Where I expect to see settlement is with the big players who either have a large amount of content or content that is particularly valuable. The New York Times may end up with a settlement and licensing agreement, possibly in which OpenAI pays to use the New York Times content.
There’s enough money at stake that we’re probably going to get at least some judgment that sets the parameters. Class-action plaintiffs, I think have stars in their eyes. There are a lot of class actions, and my guess is that the defendants will fight them and hope to win on summary judgment. It is unclear whether they will go to trial. In the Supreme Court Google vs. Oracle The case pushes fair-use law very strongly toward resolution on summary judgment, not before a jury. I think AI companies will try very hard to get these cases decided on summary judgment.
Why would it be better for them to win on summary judgment vs. jury verdict?
It’s faster and cheaper than going to trial. And AI companies are worried that they won’t be seen as popular, which many people are going to think, Oh, you made a copy of the work that should be invalid And don’t dig into the details of the fair use doctrine.
There are lots of deals between AI companies and Media outletsContent providers, and other rights holders. Most of the time, these deals seem to be more about exploration than foundational models, or at least that’s how it’s described to me. In your opinion, would licensing content for use in AI search engines—where answers are obtained through Retrieval Augmented Generation, or RAG—be something that is legally binding? Why are they doing it this way?
If you use extended generation of retrieval on targeted, specific content, your fair-use argument becomes more challenging. It’s much more likely that AI-generated search output is going to produce text taken directly from a specific source, and much less likely to be a fair use. I mean, it is can Maybe—but the risk area is that it’s much more likely to compete with the original source material. If, instead of directing people to a New York Times story, I give them my AI prompt that uses RAG to take the text directly from that New York Times story, that seems like a replacement that could hurt the New York Times. Legal risks are high for AI companies.
What do you want to know about generative AI copyright battles that they may not know, or have been misinformed about?
The thing I often hear mistaken as a technical matter is the idea that these are just theft machines. All they’re doing is taking my stuff and then mashing it back up in the form of texts and responses. I’ve heard a lot of artists say it, and I’ve heard a lot of normal people say it, and it’s not technically correct. You can decide whether generative AI is good or bad. You can decide it is legal or illegal. But it’s really something fundamentally new that we haven’t experienced before. It has to be trained on a bunch of content to understand how sentences work, how arguments work, and different facts about the world. It doesn’t mean it’s just copying and pasting things or making a collage. It’s really creating things that no one could have expected or predicted, and it’s giving us a lot of new content. I think it is important and valuable.
