Preventing problems like this seems to me to be what AI as we know it is least suited to - failures of human organizations generally seem to be failures to choose the right context due to no individual grasping it, and my impression of AI today is that it doesn't even engage with the problem. You can have a computer program that recognizes cats, or plays Go, but nobody even thinks about "how do we make this same program spontaneously respond to a spilt glass of milk, or a hostage situation, or a fire alarm...", let alone take into account everything in the world while doing so. It feels to me like people have tunnel vision that is getting worse, and the "intelligent" software is inheriting that.
