Research
156d agoLLMs Expose Online Anonymity
Researchers demonstrate that large language models can reverse engineer user identities from seemingly innocuous text, exposing the limits of online privacy. By correlating linguistic fingerprints with public data, the study shows that even pseudonymous posts can be linked to real-world identities.
The Signal
This raises concerns for social platforms, regulators, and the design of privacy‑preserving AI systems worldwide, echoing OpenAI and Anthropic.