This is a common refrain from fans of all breeds. I’m not sure that I buy it. The null hypothesis is that the “proper stimulation” correlation with “propensity to attack” is a function of population prevalence. There are simply more dogs that don’t receive the “proper stimulation” than dogs that do (especially with True Scotsman-style goalpost moving) so they are overrepresented in attacks. But anecdotally we’ve seen the testimony across all breeds - “my dog is so gentle and likes to play with children, I can’t imagine where this attack came from”.

I’d have to have some serious data on non-attack-involved dogs before I’d consider rejecting the null hypothesis.

In any sort of analytic consideration here the first thing to look at is whether the breed is capable of causing grievous harm. Temperament second, and treatment/training third. And of course overall prevalence. But in terms of the policy decision around breed-specific legislation all that really matters is the number of attacks.

I'd see a license for breeding and owning GSDs. They're not ill-socialized like a lot of other attack-prone breeds are, but they do need to herd and run around and be stimulated. It's literally in them at an atomic level to do so, and you can't simply train it out of them.

So many dogs are just ill-suited for the modern concept of having a dog. This thing where we just have a dog around a <1500sqft home where they eat and sleep and occasionally go for a walk is less than 100 years old. For the other 20,000+ years of hominid-canine interaction, they had an actual job to do that was typically physically and mentally demanding. They were warding off wolves or enemy tribes, or helping us hunt, or pulling loads. Now they just sit around all day in this tiny little slice of territory. People get in love with the aesthetics of the breed instead of what they were actually bred to do, and it has consequences.