I am still pretty firmly of the belief that "difficulty/cost of generation" has not been the bottleneck on "disinformation" for quite some time, and also, quality of disinformation has not really been that important. Yes, AI might make "better" disinformation cheaper and easier to produce. But I don't think that those are the qualities that are holding the disinformation problem back from getting bigger, so I don't really expect AI to be able to make the problem much worse.