But the chatbots themselves could very well be violating copyright laws on a massive scale. They must have used almost anything ever written as input, and I bet they didn't get permission from millions of different authors to reproduce it.
it's pretty much confirmed at this point that every AI company that has its own model has violated copyright laws in some way to train it.
facebook for example, had court documents confirming that they torrented a shit ton of books to train their Llama model.
https://www.wired.com/story/new-documents-unredacted-meta-copyright-ai-lawsuit/