No matter how safe your robotaxi, you are subject to the risks of the least-safe-driver sharing the road. Even a perfect robotaxi would get bashed by human drivers roughly once every 2 million miles. At a certain safety threshold, robotaxi platforms will be able to better risk-reduce transportation by making choices that push miles-share rather than further reducing at-fault accidents. Imagine the platform operator ships an AI update that can operate more safely at the same speed or as safe at a faster speed. Beyond a certain threshold, running the vehicles faster will reduce aggregate accidents further than would directly reducing the accident rate of the robotaxi itself.