Red Team/Blue Team: AI Agents as adversaries for building fraud models

There are countless stories of fraudsters using AI to social engineer or directly defraud companies. Feels obvious that AI can be used to harden systems/defenses. I just don't see that much of it.