I remember less than a year ago thinking (and posting) the same thing. Until you actually think about it, it does appear from the outside, to be a relatively easily solvable problem.
I still believe It will have to be done, one way or another, though I have no idea how it will be done.
For a simple, dump system it genuinely is easily solvable. The problems you're thinking about mainly crop up with smart systems, or anything self-improving.
"Even a non-anthropomorphic human intelligence still could pose threats to mankind, but they are probably manageable threats. The greatest problem is that such artificial intelligence may be indifferent to human welfare. Thus, for instance, unless otherwise programmed, it could solve problems in ways that could lead to harm against humans. But indifference, rather than innate malevolence, is much more easily cured. Artificial intelligence can be programmed to weigh human values in its decision making. The key will be to assure such programming."
More at:
http://www.businessinsider.com/robot-apocalypse-2014-7