Sure, but would that be valuable? Non-benevolent takeovers help us fine-tune controls by noting new potential ways that we might not have thought about before, this can help in ai safety research and policy making. I'm not worried about 'accidentally' losing control of the planet to a well-meaning AI.