[ggreer]

If you look at the NHTSA's data[1], certain manufacturers have strangely low numbers of reports, and some manufacturers seem to be missing from the data entirely.

If you read the NHTSA's standing order[2], it says:

> Crashes that meet specified criteria must first be reported within one or five calendar days after the manufacturer or operator receives notice of the crash, and other ADS crashes must be reported on a monthly basis.

The phrase, "receives notice of the crash" is very important. I'm betting that most manufacturers have low report counts because only a few (like Waymo and Tesla) have instrumentation on their vehicles that automatically notify them when a crash has occurred and whether that crash involved self-driving software within the previous 30 seconds. Everyone else has to get notices via media reports, lawsuits, their own internal testing, etc. Even if a manufacturer has OnStar in their cars, it doesn't look like OnStar automatically reports the crash to the manufacturer, just emergency services and insurance companies. If they do, they'd also need to send info about the usage of any self-driving software before/during the crash.

The order seems well-intended, but it does have the side effect of discouraging companies from improving their data collection.

1. https://www.nhtsa.gov/laws-regulations/standing-general-orde... Look at the ADS and ADAS dashboards, and go to the "Reporting Entity" tab.

2. https://www.nhtsa.gov/sites/nhtsa.gov/files/2023-04/Second-A...

[FireBeyond]

> I'm betting that most manufacturers have low report counts because only a few (like Waymo and Tesla) have instrumentation on their vehicles that automatically notify them when a crash has occurred

Tesla doesn't count fatal accidents in its normal reporting.

It also doesn't consider any incident where there was no airbag deployment to be an accident. Sounds potentially reasonable until you consider:

- first gen airbag systems were primitive: collision exceeds threshold, deploy. Currently, vehicle safety systems consider duration of impact, speeds, G-forces, amount of intrusion, angle of collision, and a multitude of other factors before deciding what, if any, systems to fire (seatbelt tensioners, airbags, etc.) So hit something at 30mph with the right variables? Tesla: "this is not an accident".

- Tesla also does not consider "incident was so catastrophic that airbags COULD NOT deploy*" to be an accident, because "airbags didn't deploy". This umbrella could also include egregious, "systems failed to deploy for any reason up to and including poor assembly line quality control", as also not an accident and also "not counted".

[Veserv]

There are really two things at play here.

First, every manufacturer should be required to collect and produce scientifically rigorous evidence of safety for these systems. Any system that does not increase danger, i.e. within the range of human safety, would generate a easily manageable number of incidents to characterize. The total number of US traffic fatalities is only on the order of 40,000/year. Even if it took a staggering 10 hours of 300$/hr lawyer time just to characterize each incident, that would amount to a mere 120 M$ for the entire US. That is just 50 measly cents per car annually. And we are talking literal deaths here. Not doing so is obscene.

Literally every manufacturer of ADAS systems fails by this metric. None of them is acceptable. Arguments about who is failing double turbo hard is a meaningless exercise. The minimal standard they should be held to is clear and all of them should need to meet it or they should pull their systems from the field until they feel like taking human lives seriously.

Second, although all manufacturers fail at the standard that they should be held to and should be censured for that, only Tesla actively and deceptively markets their “safety performance” according to these very same metrics they claim are misleading. The reported data objectively undercounts incidents. Despite this, Tesla intentionally and deceptively markets the data as the true rate of incidents to overstate their safety. No other manufacturer does this as far as I am aware.

Every other manufacturer just relies on the prevailing wind that scientifically rigorous safety evidence collection is not needed when deploying new features. But none intentionally misrepresent the absence of evidence of danger as evidence of safety. That is pure scientific and business malfeasance that Tesla is guilty of and, even barring a change in evidence collection requirements, should be punished for.