From my previous comment, it looks like NHTSA is moving faster than I predicted. We're now at step 1, with this Advance Notice of Proposed Rulemaking.
(edit: I jumped the gun, we're still at step '0' on my original list)
Most of this notice seems to be a report on why 'impaired driving' is bad. I see alcohol, cannabis, mobile phone use, drowsiness...etc.
Due to technology immaturity and a lack of testing protocols, drugged driving is not being considered in this advance notice of proposed rulemaking.
Makes sense.
There is no clear and consistent engineering or industry definition of ‘‘impairment.’’
Yep, another unclear request by Congress.
NHTSA believes that Congress did not intend to limit NHTSA’s efforts under BIL to alcohol impairment.
Okay, that's fair.
Camera-based-systems, however, are increasingly feasible and common in vehicles.
Uh-oh...
The Safety Act also contains a ‘‘make inoperative’’ provision, which prohibits certain entities from knowingly modifying or deactivating any part of a device or element of design installed in or on a motor vehicle in compliance with an applicable FMVSS. Those entities include vehicle manufacturers, distributors, dealers, rental companies, and repair businesses. Notably, the make inoperative prohibition does not apply to individual vehicle owners. While NHTSA encourages individual vehicle owners not to degrade the safety of their vehicles or equipment by removing, modifying, or deactivating a safety system, the Safety Act does not prohibit them from doing so. This creates a potential source of issues for solutions that lack consumer acceptance, since individual owners would not be prohibited by Federal law from removing or modifying those systems (i.e., using defeat mechanisms).
Note that "make inoperative" does not apply to a "kill switch" in this case. NHTSA uses the term to mean "disabling required safety devices". For example, as an individual vehicle owner, it's perfectly legal for you to remove the seatbelts from your car, despite Federal requirements. But it's illegal for the entities listed above to do it. (This example doesn't extend to state regulations. It's legal for you to remove your seatbelts, but may still be illegal to drive a car without them.)
There's a short 'discussion' here regarding how to passively detect impaired driving, noting the difficulties of creating such a system. Followed by a note that basically says if they can't do it within 10 years, NHTSA can give up and not do it, as stated in the Infrastructure law.
There's a long section on how to detect various types of impairment, current methods of preventing impaired driving, etc. An interesting section about detecting blood-alcohol level using infrared sensors embedded in the steering wheel. Body posture sensors can be used to detect driver distraction.
This is followed by a brief overview of the technologies NHTSA is considering:
Camera-Based Driver Monitoring Sensors
Hands-On-Wheel Sensors
Lane Departure and Steering Sensors
Speed/Braking Sensors
Time-Based Sensors
Physiological Sensors
On page 850 (21 of the PDF), NHTSA asks for feedback to several questions. There are a few pages of relevant issues, so I won't cover them here. If you wish, you can go here to leave a comment. Please don't leave irrelevant garbage like "I oppose this on the grounds of my Constitutional rights..." While applicable in this situation, it's irrelevant to NHTSA, and commenting like that will just waste everybody's time. There's a section on page 855 (26 of the PDF) about Privacy and Security.
That's that. Let me know I can answer any of your questions. I'll try to come back to this post throughout the day and see what's happening. But, I do not work for NHTSA, so can't remark on agency thought process.
One of the things that was posited was that cars would look at the way you were driving, and if you were driving "erratically" it would shut off.
So what happens when you're trying to get someone to a hospital because they've been seriously injured and are bleeding to death in your car? No, it doesn't happen very often. But I can think of at least one case: Kentucky Ballistics, who had a rifle explode and blew shrapnel into his jugular. You will absolutely be driving erratically in those circumstances; exceeding the speed limit, weaving, honking, turning without signaling...
On the other hand I can think of many more cases when someone was killed by erratic driver. It kind of sounds like you're ignoring something occurring every day and focusing on fantasy scenario. It's obvious that eliminating even 1% of accidents caused by erratic drivers would save more lives than people racing to hospitals do.
I'm not saying that shutting off cars based on some AI analyzing your driving patterns is a good idea but you really need to think about another argument than "this one guy would probably die this one time".
Also, that's what ambulances are for. Fixing the ambulance service would be a better idea than hoping people will manage to race to hospitals without killing anyone.
Many people have died waiting for ambulances that didn't come, or took too long. Houses in my area have burned down because the fire department couldn't figure out where an address was because GPS gives them the wrong location.
If I slip while working with a chainsaw and cut my femoral artery, I'm not going to tell my wife to wait for an ambulance; I'm going to get a tourniquet on and have her drive me as fast as she can to the hospital, because that will save me 20 minutes--minimum, and that's if they're not already out on a call on the other side of the county--over waiting for an ambulance.