But then if the AI is detected that person can be permanently banned. No more AI. No new accounts.

So if someone compromises your identity they can unperson you? How will the AI be detected? By another AI?

"So if someone compromises your identity they can unperson you?"

You've identified a problem that unrelated systems also have. Like banks and identity theft. This solution isn't responsible for causing that problem.

"How will the AI be detected? By another AI?"

However a platform likes to. Let the best platform win.