One exception is that UMG v. Anthropology This is the case, because at least early on, previous versions of Anthropic would generate lyrics for songs in output. That's a problem. The current state of the case is that they have put in place safeguards to try 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 looking. a preliminary injunction.
Ultimately, the more difficult question for AI companies is not Is it legal to participate in training? Its What do you do when your AI produces results that are too similar to a particular work?
Do you expect the majority of these cases to go to trial or do you see settlements coming?
There will probably be some settlements. Where I expect to see settlements is with large players, those with large amounts of content or particularly valuable content. The New York Times could end up with a settlement and a licensing agreement, perhaps paying OpenAI to use New York Times content.
There's enough money at stake that we'll probably get at least some judgments that establish the parameters. The class action plaintiffs, I have a feeling they have stars in their eyes. There are a lot of class action lawsuits, and my guess is that the defendants will fight those lawsuits and hope to win at first instance. It's not clear they will go to trial. Supreme Court in Google and Oracle the case pushed fair use law very strongly in the direction of being resolved by summary judgment rather than before a jury. I think AI companies will do their best to resolve those cases by summary judgment.
Why would they win on summary judgment instead of a jury verdict?
It's faster and cheaper than trying it out. And AI companies are worried that they won't be seen as popular, which many people would think, Oh, you copied work that should have been illegal and does not go into the details of the fair use doctrine.
There have been a lot of deals between AI and companies mediacontent providers and other rights owners. Most of the time, these trades seem to be more about searches than fundamental patterns or at least that's how it's described to me. In your opinion, is licensing content used in AI search engines—where answers are derived from retrieval augmented generation or RAG—a must in terms of legal? Why do they do it this way?
If you're using retrieval enhancement on specific, targeted content, your fair use argument will be more challenging. It's more likely that an AI-generated search will produce text taken directly from a specific source in the output, and that's less likely to be fair use. I mean, it maybe okay—but the risk is that it's more likely to compete with the original source material. If instead of directing people to a New York Times story, I give them an AI prompt that uses RAG to pull text straight out of the New York Times story, that seems like a viable alternative. could harm the New York Times. Legal risks are greater for AI companies.
What do you want people to know about general AI copyright struggles that they may not know or may have been misinformed about?
The thing I most often hear that is technically wrong is the notion that these are just plagiarism machines. All they're doing is taking my content and then mashing it up as text and responses. I've heard a lot of artists say that, and I've heard a lot of normal people say that, and it's not technically true. You can decide whether Generative AI is good or bad. You can decide whether it's legal or illegal. But it's really something fundamentally new that we've never experienced before. The fact that it takes training a bunch of content to understand how sentences work, how arguments work, and to understand various facts about the world doesn't mean it's just copy-and-paste. everything or create collages. It's truly doing things that no one could have expected or predicted, and it's giving us a ton of new content. I think that's important and valuable.