Tesla's system are unsafe, by default, if they don't follow safety life cycle. And they don't - I saw dick-sharing apps that had better life cycle processes.
We are lucky that our ancestors were not so risk averse, because if they were we would not have cars at all, or airplanes.
That makes sense to me. I'm happy to accept the presence of full self-driving technology on the roads, for other people, once it has an accident rate comparable to or slightly better than humans.
I personally won't use one until it is so much safer than a human that the level of safety is the #1 feature though. Until then, what's the upside? I can screw around on my phone more often? I do that too much already, and it's not the kind of benefit that cancels out the potential downside of "...but you died because of an unhandled edge case in version 27.1.828 of our software that you as an attentive human would have easily handled, which was fixed in the next release" which just seems like such a banal way to go for more screen time.
I don't think my take is drastically out of the mainstream. It also seems to me the main thing separating my point of view from "ban all FSD until perfect" is a willingness to let other people make choices I don't think are good.
We can’t expect this technology to work well in the real world unless it is tested in the real world before it works well.