Neural networks (or any learned model) may often behave differently from the intentions of the programmers. A useful distinction between the failure modes is between outer and inner misalignment.