That's exactly the problem, isn't it?

AWS is not responsible for what you do with your data. AWS, on its own, is not going to make backups for you so you can recover from a failed DB migration you did yourself. AWS cannot be held responsible for your admin fat-fingering and deleting the entire prod environment.

However, AWS is responsible for the underlying infrastructure. The website you linked clearly shows "storage" as falling under AWS's responsibility: your virtual hard drives aren't supposed to just magically disappear!

If they can just nuke your entire setup with an "Oops, sorry!", what's all the talk about redundancy and reliability supposed to be worth? At that point, how are they any different from Joe's Discount Basement Hosting?