What's up with model collapse?

I'm amazed by how much of the internet is clearly AI generated now. Everything from random articles to youtube videos to instagram images/reels, it's everywhere. Overall, I'd say it has clearly lowered the quality of data out there, and increased the noise in a difficult to filter way. The idea of model collapse is that this process is kind of poisoning the training data for the same models that are generating all that crap. Is this a real issue? Are the big companies investing a lot on cleaning the AI slop