But that breaks OpenAI’s (et al) entire business model. Those AI companies can barely afford to operate as it is, while they scrape the entire web for free. I don’t see how they could keep above water once every website starts paywalling their stuff.

Is it the training or running costs that are putting them in debt? Presumably we'll eventually get to a point where the models are good and they can train less and charge enough to turn a profit. Maybe then they can do a revshare with content creators. Maybe something like YouTube

They'll have to figure it out because people aren't gonna write content just for openai to scrape it, the economics don't work.