Password researchers found LLM-generated fixes often overlooked root causes, architectural context, and long-term security ...