while the avengers might never have to face Chat-GPT, there are some interesting things to come out of that farce. Once AI mimicry becomes sufficiently matched with a complex enough "will" to find an answer, AI directed code will slither through the cracks in some inappropriately connected system (biotech? financial? utilities infrastructure? automated weaponry?) to cause pain for whoever the meat puppet that posed the question has decided is an enemy.
And stopping research? LOL. Wow. That's really working well for nuclear proliferation..