AI Risk Brief: Trust-Chaffing Failure Modes in AI Systems and Downstream Public Harm