LLMs can deanonymize internet users at scale
A recent publication by the SPY Lab ask how good large language models are at re-identifying anonymous online users. This is a collaboration between Daniel Paleka, Joshua Swanson, Michael Aerni and Florian Tramèr from ETH Zurich’s Department of Computer Science (D-INFK), together with Simon Lermen (MATS) and Nicholas Carlini (Anthropic). Although deanonymization is a decades-old