That’s not touching the ethics of stealing everyone’s work for your AI then selling it back to them.

Yeah, it’s the “selling it back to them” part that gets me. I don’t want to outlaw scraping, because then the Wayback Machine couldn’t function. The Wayback Machine is free, though. It’s a public service. (One which is very much needed, IMO.) If, instead, the Internet Archive required that you pay for a subscription in order to use it, I think most people would be very upset, and rightfully so.

You want to scrape the Web to feed your LLM? Cool! Make it an open-weight model.