It is well known that A.I. companies have red-teaming, RLHF, and guardrail teams specialized in protecting against hate speech, bomb-making, or any crazy idea that pops into a person’s destructive mind. However, even with protection, the A.I. can be “tricked” and bypass the blockade, creating monstrosities. If there are psychopathic humans capable of manipulating and convincing crowds, just imagine a tool built with data from the entire internet.
