goodLongRead

Not saying whether the conclusion is right or wrong here, but this isn't a "study" in the sense of something published in a rigorously peer-reviewed research journal. It's a random car-ranking company scraping data from the Federal Highway Administration and posting it on their own website. The source is pretty low quality.

Again, just because it's low quality, doesn't mean it's wrong, but I'm not regarding this as very authoritative evidence.

As someone typing this while driving a Tesla on autopilot I understand

Well, they had good safety rating IIRC - but no amount of safety measures is going to stop idiots from doing idiotic things with an explosive on wheels.