"What about the financial / legal / social risk of your AI agent doing something bad?" the same way we treat it like human making mistake??? AI cant code thems

"What about the financial / legal / social risk of your AI agent doing something bad?" the same way we treat it like human making mistake??? AI cant code themselves, someone command them to create something