I guess it's true an AI could pose an existential threat to humanity, not only because it might see us as a threat, but also because it might not even share human values at all. However, I don't know how he plans for those regulations to come about, or even be effectively implemented. I mean, I think we can all agree that if an AI can be created, someone is going to create it sooner or later. Probably the military is going to do it first, considering the benefits it could potentially bring to them, but the resources to do so will eventually be at the disposal of a lot more people as technology improves and access to it increases. Or, who knows, maybe it will be created accidentally like in ghost in the shell.

But I do agree with him on the rest: once someone creates it, there will likely be no way to control it.