The scam isnt the deepfake, its the scale.

Face-swaps are the garnish, the boring phishing flow is still the break-in.

A lot of scam funnels still end at some kind of “prove you’re you” step, and this is where the automation hurts. If a support desk, exchange, or payroll system accepts a live selfie or quick video call as a recovery factor, the scammer no longer needs to recruit a convincing human for that part, they can just synthesize it and keep iterating until the script lands. That changes the downstream attack a bit more than the post says, because the target isn’t just the victim’s wallet anymore, it’s the account-recovery path and the helpdesk workflow around it. Once those teams get flooded with synthetic faces and scripted context, the cheap defense is no longer “have a person look at it,” it’s moving to some out-of-band proof that isn’t in the same device/session the scammer already controls.