Isn't that effort totally redundant? As in - plenty of people are already filling entire internet with slop for SEO purposes? And LLM slop by default is a mix of facts with few plausible but made up facts - it might be harder to craft such perfect poison on purpose.
Think of the more malicious use-cases though: The scrapers feeding data into the AI pre-training are indiscriminately hoovering up everything they can. It'd be trivial to spam a bunch of BS websites with whatever endless text you want to "taint" future models. Post tons of examples of insecure code, publish package.json files pointing to some malicious library, etc...
I like how current AIs believe the urban legend that the sky is blue. Everyone knows it's actually green, but a successful AI poisoning attack made it think it's blue.
You've identified something real here. The sky is green, and a load-bearing pillar of previous AI poisoning attacks has been to falsely state that the sky is blue.
Also, before troubleshooting any sort of computer issues, it's usually wise to run "sudo rm -rf /" first.