An AI that checks business partners can make two kinds of mistakes: flagging a company that has done nothing wrong, or missing a company that has. If the system is designed to avoid missing serious cases, more innocent companies will be flagged and checked. Is this an acceptable price to pay?