>These days, anyone can give an LLM footage from any source, and get this kind of information.

Is a bit misleading itself, to do this at scale requires all those iffy data centers.