Safety First for Automated Driving (Intel)

Anyone have thoughts on this?
seems not fully intelligible or practical.

The producers of this document need a framework for convincing regulators that their product should be approved prior to the availability of useful operational statistics. Ideally they also want to set a liability expectation that a system is presumed safe and is not liable for damages so long as it performs according to the design requirements approved by regulators. I think this is the start of a conversation with regulators and society regarding what those standards would need to be.