I get really concerned about AI, only for the ethical issues that inherent to it. At its core, AI must be programmed - and that programming may have long term consequences out in the wild.
Think of a self driving car. Self driving cars are becoming more and more aware of their surroundings, then make decisions based on that data. Consider this example:
A self driving car gets cut off by a car on a narrow street, and has to make a decision on what to do. A group of nuns are on the sidewalk to the left, and group of school children are on the sidewalk to the right. A school bus is right next to it. The car has to make a choice (albeit a horrible choise) on what to run into. At some point, those "ethical" decisions have be thought out during coding.