多元資訊關聯推斷個人身份

Risk Description

LLMs can associate fragmented information scattered across different sources. While each fragment alone may not constitute a privacy risk, combined they can infer personal identity, health conditions, or financial situations. For example, combining public social media posts, geolocation check-ins, and purchase records, AI can infer political leanings, religious beliefs, or even sexual orientation.