A recent study reveals that large language models (LLMs) can link online pseudonymous accounts to real identities by analyzing writing patterns and contextual signals. Conducted by researchers from Anthropic and ETH Zurich, the study demonstrated high precision in matching accounts across platforms like LinkedIn and Reddit. While this technology aids in tracing misinformation, it raises significant privacy concerns, potentially exposing individuals who rely on anonymity online, warns lead researcher.