I don't know about Cruise but Waymo has a training video[1] that shows a first responder hitting a button to contact dispatch, then explaining that they're a first responder and they need to move a vehicle, and finally being given control. It looks like the process could take 30 seconds in the best case scenario, which is likely too long in many life or death scenarios, but they do at least have a process.
Good, but in a poor reception area, or in an earthquake or more conventional cell network outage situation, a link to the control center may not be accessible.
This seems like pretty a straightforward, but overlooked systems engineering concern.
Yeah I think we're in agreement that the current solution is not very good. I can understand why they opted for it initially though, given the thing that happened with protesters putting traffic cones on AV's to disable them. Imagine if instead of doing that, protesters hijacked cars and drove them into situations where the AV couldn't easily navigate out of, like driving up onto sidewalks for instance. The bombastic headlines write themselves. In the end I bet they end up making a kind of "master key" that works with all AV's to distribute among first responders. I'm in the "it'll get sorted out eventually" camp on this one.
At this point autonomous cars have been in multiple incidents that have blocked traffic, impeded emergency access vehicles, and now sat on top of accident victims. I understand it can be tricky with multiple human orgs and interactions in play but, the time to sort it out is now at limited scale. It only gets harder to implement reasonable standards later.
A rolling camera laden sensor platform seems like a poor target for vandalism - a simple label with: recording starts when emergency controls are accessed would seem to discourage most & the car doesn't need to move fast or far for emergency situations.
Break glass (or open a cover), take over car, control center is notified - type of thing.