If it had 5 errors 4 of which were not deadly due to random circumstances like no vehicle with right of way passing, then that is worse than if it committed one error that was deadly. I might fix the second implementation and scrap the first.
Attacks that are truly limited to one company and not other companies with overlapping community, philosophy or providers seems so low to me that I wonder why anyone looking at highly customized attacks hasn't considered more physical/manual means.
I don't really get the description going through a stop sign and hitting pedestrians next to a stopped car. It sounds like 2 entirely independent errors in succession as no right of way, etc, is really affected by the stop?
I thought we were talking about NIH, H is for here. If you can't scope your requirements then obviously you need to import something a bit larger than the whole world to solve the whole world's problems.
Of course these questions are hard and if you are asking them you may need a general db.
The whole industry is wrapped up in these bizarre ideas that for the slightest requirement they are going to add just one more abstraction layer and it will do everything and the kitchen sink. But sometimes the requirement is simple and doesn't deserve solutions to anticipated problems and other times it does and is worth looking for an existing solution. Rarely the problem is complex and deserves a custom solution and then one should probably make a new startup for the one yak shave, etc..
It may have been the lower hanging fruit. But they already have all the tools and behavior analysis is no longer the competitive edge. They can ask for your smoking status and deny your claim for fraud later, but (any genetic data on behavior is an update to fraud investigation value and) they can't know your propensity to non-smoker cancers.
If it had 5 errors 4 of which were not deadly due to random circumstances like no vehicle with right of way passing, then that is worse than if it committed one error that was deadly. I might fix the second implementation and scrap the first.