The thing is, it shouldn't take much in the way of "AI" to recognize that a company that has already been doing business for several months or even years is probably not going to wake up one morning and start defrauding people. Criminals are lazy, and running a business is a lot of work. It would take about 30 seconds' worth of review time on the part of a moderately low-paid staffer at PayPal to avoid most of these PayPal Media Debacle of the Week stories.

I simply cannot believe that it's that hard to distinguish between a fraudulent user and a real one, given the presence of a significant transaction history. New accounts opened by people with no discernible history? Yes, they should freeze/ban/lock first and ask questions later. Accounts that are clearly used as part of a business? Give the customer the benefit of the doubt, or at least a 5-minute phone call.