I would watch for signs that one of the big companies deprioritizes internal research. I think it will be a lagging indicator that the self driving startup scene has cooled and they think that outside startup challengers will be underfunded and easy to buy or compete with. Right now it is accelerating in the opposite direction, but I don’t see why it has to if we don’t see progress in the next couple of years.
Of course they could have secret agreements that the governments don't know about, but cartels like this are a dangerous game that tend to backfire on the members. Each one that agrees has to play along, but if any one decides to work in secret they are first and get to scoop the competition and win. Since the agreement is already illegal under most national laws there is nothing that the others can do about the cheater.
What the safety reality of self driving will be is of course an open question. The risk that someone else gets there and proves they are safer is too high for the big auto companies to ignore. Regulators already insist that all companies have any useful safety device when the device proves itself.
Note that the car companies are interconnected. The ones that are not investing in this technology have cross agreements with those that are.