I just wonder if they’ll get out of it because LLMs do reword the information instead of spitting it back out verbatim. It’s the same reason I think the image generators are safe from copyright law - it’s just different enough that they could plausibly convince a judge with a fair use argument.

What bothers me even more is all the text they had to scrape to create ChatGPT… That seems like a novel problem for the legal system because you know there’s no way they paid for all of it.