But is it really risk analysis if any negative eventuality, no matter how unlikely or infrequent, is a deal-breaker? Risk analysis is exactly that - figuring out what the risks are, their likelihoods and potential impacts, and then figuring out which ones are worth taking and which are not.

So the salient fact remains the same: deaths with autonomous driving are less than deaths without it. Isn't it unethical to prevent such a feature from going out into the world when it could be saving lives, due to concerns about some unlikely / infrequent eventuality or driver error? Is it sound risk analysis to allow hundreds, thousands of people to die when you have a feature that could save their lives?

Cars are inherently unsafe, and yet we let people drive them. Clearly we already (ethically) accept some risk in our products.