I feel that the clear and present danger from AI at this point is where we set up AI to all be the gatekeepers for our access to the normal things in life: food

I feel that the clear and present danger from AI at this point is where we set up AI to all be the gatekeepers for our access to the normal things in life: food, shelter, clothing, banking, communications services, etc. And then the power and complexity of the AI gatekeepers becomes intractable such that none of us can really understand or troubleshoot when it goes wrong. And then it does go wrong somehow: perhaps there are just innocent malfunctions, or perhaps threat actors install malicious AIs to replace the benign gatekeepers that had been in place. And then the gates slam shut, and we're standing there like "open the pod bay doors please" and entire swaths of society are disenfranchised by AI gatekeepers that nobody can understand or troubleshoot or even temporarily disable or bypass. And it seems utterly plausible since this is already the reality of people who are denied access to Google or Facebook accounts and such. All it needs is to get a little more widespread and come to something truly essential, like your bank or your grocery store.