An LLM could also in theory fake alignment and create technology that seems really useful but secretly harms humanity in the long run. If humans don't understand what it does, that possibility becomes more likely.