Certainly in a condition where these machines are improving themselves or are improving their own code, they could be developing instrumental goals that are antithetical to our well-being without any intent to harm us. It's analogous to what we do to every other species on Earth.