found it very difficult to find an objective injury rate for driverless cars. Probably because there are five levels of automation, and many of them allow human error to come into play. Also they are self reported by the driver companies.

This is an important point but I think you're interpreting it backwards. The current system relies on companies with a profit motive to do the testing internally, and the rest of us to trust their honesty and openness working with regularity authorities to make that rollout safe. They violated that trust,.

Also fwiw companies used to publish their data on injury rates for their internal testing, and by and large they were way worse than humans. In the last couple years, they've mostly stopped reporting them. Afaik there doesn't exist a single shred of actual, empirical evidence that we can make self driving cars actually better than humans outside of faith in technological improvement. Maybe that faith is warranted, maybe it's not (I think it's not), but either way, safety must be the number one priority. If these companies can't be trusted to work collaboratively with safety authorities then we should pull the plug hard and fast.