There is an exception to that UMG v. Anthropic In this case, because at least early on, previous versions of Anthropic generated lyrics for songs in the output. This is a problem. The current state of that case is that they’ve put in place measures to prevent that from happening and the parties have sort of agreed that until the case is settled, those guarantees are enough, so they’re not looking for it anymore. Preliminary order
In the end, the harder question for AI companies isn’t this Is participation in education legal? This is What do you do when your AI produces output very similar to a specific task?
Do you expect the majority of these cases to go to court or do you see settlements on the horizon?
There may be some settlements. Where I expect to see settlements is with the big players who either have a lot of content or valuable content. The New York Times may end up with an agreement and a licensing agreement, perhaps where OpenAI pays to use The New York Times content.
There is enough money at stake that we will probably get at least judgments that set the parameters. The class action plaintiffs, I sense, have stars in their eyes. There are many class actions, and my guess is that the defendants will resist them and hope to win on summary judgment. It is unclear if they will go to court. Supreme Court in Google vs. Oracle The case narrowed the Fair Use Act heavily toward a judgment settlement, not before a jury. I think the 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 versus a jury verdict?
It’s faster and cheaper than testing. And AI companies worry about not being seen as popular, and many people think, Oh, you made a copy of the work that should be illegal And don’t go into the details of the fair use doctrine.
There have been many deals between AI companies and media, content providers and other rights holders. Often times, these deals seem to be more about search than underlying models, or at least that’s how it’s described to me. Do you think licensing of content used in AI search engines – where answers are generated through retrieval augmented generation, or RAG – is something that is legally mandated? Why do they do this?
If you’re using next-gen retrieval on targeted, specific content, your fair use argument becomes more challenging. An AI-generated search is much more likely to output text directly from a specific source, and much less likely to be fair use. I mean it was able be – but the danger zone is that the possibility of competing with the original source is much higher. If, instead of directing people to the New York Times story, I instruct my AI that uses RAG to extract the text directly from the New York Times story, that seems like an alternative that could hurt the New York Times. . The legal risk is higher for the AI company.
Want people to know about AI copyright litigation they may not have known about, or been misinformed about?
What I often hear mistaken as a technical matter is the notion that these are just plagiarism machines. All they do is take my stuff and then chop it up into text and responses. I hear a lot of artists say that, and I hear a lot of lay people say that, and it’s just not true as a technical matter. You get to decide whether the generative AI is good or bad. You can decide whether it is legal or illegal. But it’s really something new that we haven’t experienced before. The fact that understanding how sentences work, how arguments work, and understanding different facts about the world requires teaching on a bunch of content doesn’t mean it’s just a matter of copying and pasting things or creating a collage. It really produces things that no one can expect or predict, and gives us a lot of new content. I think this is important and valuable.
