Comment by skywhopper

It was a proof of concept and one intended to cause as little collateral damage as possible. But if Google's AI can't tell the difference between a little joke and something real (and of course, it can't, and never will be able to do so), that's a weakness that can be exploited both on a bigger scale and more subtly.

If you don't think bad actors are already attempting this sort of thing (and have been, ever moreso the past four years, including with the help of the very LLM tools they are trying to subvert!) and learning how to manipulate these systems, you are being naive.