falling down and breaking things?

I do think robotics would come up with more safety mechanisms (provably safe motion planning etc) just because the risk is a lot more serious than LLMs spitting half-truths