There’s always false negatives for systems like this, and if Stripe wrongly tells you someone is good then are you liable for acting on that decision?
And I don’t think anyone builds their own KYC from ground up - it’s more whether you have context for a decision with your own vendor implementation vs a black box yes or no.
EDIT: For those that think that just opening the accounts as Stripe would be a workaround, the answer to that is "beneficial ownership" and is part of KYC.
They're not the treasury, they're the link that doesn't require me to be a huge entity with bargaining power to convince a bank to partner with me.
To the false negatives problem itself though, I think your implicit assumption is the correct one - that false negatives are a bigger potential problem than false positives. In my experience, the false positive rate on declines was immaterial to the business. The impact of a false negative is definitely a different question.
Yes, they are black boxes, but they work pretty well considering banks all over the world use them.