Yeah, I think this is a bigger threat, at least for the moment. It's not automated systems causing havoc in themselves, it's hostile actors using tech to have outsized impacts on society.
As you point out, AI lets criminals and hostile regimes get access to a level of technical knowledge that was previously reserved for larger and more well funded organisations and nation states, that required experts with years of experience to understand. Things like drones and missiles are available to people that would otherwise be incapable of building them.
And it's not just limited to that really. Online scams and fraud are arguably easier to set up than ever, given you can use AI to make fake sites for phishing services, write somewhat convincing sounding emails or impersonate the voice of a target's friend or relative to trick someone into sending money.
Plus, any weakness in a system that a sentient AI could theoretically exploit can be exploited by a criminal or hostile regime right now. For all the talk of an AI taking over a weapon system, the bigger threat is a cartel, terrorist group, cult, lone nutcase criminal or dangerous regime doing it.