The Probabilities of AI Breaking Out of Our Control

Some people are quite worried about AI going a different direction where it's not a benevolent dictator, but a malevolent dictator. The escape attempts for GPT-4 and how it was tested keep me up at night. I think that this, depending on exactly how we want to sketch this scenario, if we're talking about AI breaking out of our control and seizing weaponry or taking over the world. What do you think the probabilities of this arc is? We'll put it that way.