Auto insurance risk models have credit scores in their pricing algorithms in virtually every jurisdiction where that is permitted (California prohibits that, yet another reason why their car insurance rates are relatively high.Undercut banks on car loans. Good idea!
(Edit: Who knows, maybe driving behavior correlates to credit score, or other data mining stuff not yet explored.)
That practice probably is a net benefit for Tesla. For example in the Tesla 2018 Lease securitization the average FICO score was between 831 and 860 sample values, probably the highest in the industry (They were when I last checked some years ago. It is vanishingly rare to have such high scores in anything).
The actual auto algorithms are not really very complex because driving behavior (driving citations, accidents) is available, marital status and home ownership (high correlations),
sex (women less risky than are men), age <25>70 are the highest risks, plus address and vehicle storage (surprise! On-street parking is higher risk) All those are all readily available.
Some of the most predictive variables themselves have huge cross-correlation with each other and also with legally or morally precluded ones).
The net of all that is causing car insurance rates that are assigned based on actual driving, monitored by vehicle (Tesla), or Telematics e.g Progressive;
Telematics Devices for Car Insurance
Can a car telematics device help you save on auto insurance? Learn how the data from a telematics device could lower your premiums.
www.progressive.com
All these things are commonly used around the world with ‘no claims bonus’ schemes added which themselves correlate closely with the other factors.
Active driving behavior directly measured beats all the other solutions for speed, accuracy and precision. Those do not help with initial underwriting, though.