Given the nature of the system, when would she have been notified that she needed to take control? Additionally, what did Uber tell her about how the system operated? Ultimately it is on Uber for the lack of serious consideration. Rolling out level 2 & 3 autonomous driving systems where driver attention (especially non-experts in an open world) is still required on a wide scale is absolutely dangerous given human nature. Most people believe that you have to completely jump to level 4 & 5 autonomy which represents fully autonomous systems.

Think of it like this: "Hey, let's get this person who isn't an employee of the company, gets paid a pittance in a "gig" economy where extreme "hustle" is needed, probably isn't an attentive driver anyways and put them as a critical backstop in a system that is still a prototype and that they most likely don't understand."

The Uber engineering team lacked or ignored human factors experts.

I think it depends who was legally in control of the vehicle. I assume (and I relise it could be different, but this is just creating a model) that the safety driver would legally be considered to be in control of the vehicle, and as such responsible for the crash. She did after all have the ability to prevent the crash had she not been negligent.

I assume for instance that if I were to use Tesla AutoPilot on the road in the UK (I don't have a Tesla so I haven't looked into this) and my car crashes into someone while a) it's enabled and b) I'm not paying attention that I am still 100% at fault.

Until self-driving cars can legally be in control of themselves, absolving occupants of any responsibility, I'd assume that this is, or at least should be, the case.

I don't think Uber is clean in this to be clear, I suspect they were cutting corners to stay competitive, and I just don't trust them at all to make decisions that are in the interest of the general public, but the direct criminal responsibility seems to lie with the the safety driver, even though it seems that Uber should be sueable for something.

"legally be in control of themselves"

Never. They have no skin in the game. If they did, locking them inside a car for their life would be illegal.

At some point I suspect laws will change if self driving cars become good enough. I don't know where the liability will lie, perhaps the car manufacturers, or the insurance companies.

Indeed. The "safety drivers" are designed to protect Uber's liability, not pedestrians.

see: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2757236 ("Moral Crumple Zones")

Much of the blame in the 737 MAX crashes is aimed at failed automation.

https://www.theverge.com/2019/3/15/18267365/boeing-737-max-8...

In his defense, it's much harder to react to emergency situations if you are not already fully engaged. It's similar to the difference between dropping a breakable item versus watching someone else drop it. You will start reacting to your own folly much faster than passively watching it unfold.

Uber should have added a camera which watches the driver and checks if they're paying attention.

If not it alerts, turns the warning lights on and slowly stops the car.